literature review on magnetic resonance imaging

• Introduction
• Conclusions
• Article Information

FFR indicates fractional flow reserve; ICA, invasive coronary angiography.

Study-specific odds ratios were pooled using the Mantel-Haenszel method. The dashed vertical line represents the pooled effect estimate. Squares represent weighted point estimates of the effect of each study. The diamond size is proportional to the overall weight in the random-effects model.

P  < .001 for all comparisons. AER indicates annualized event rate; CV, cardiovascular; LGE, late gadolinium enhancement.

eMethods. Keywords Used to Identify Diagnostic and Prognostic Studies

eFigure 1. PRISMA 2020 Diagrams of Search Results

eFigure 2. Diagnostic Test Accuracy Studies Quality Assessment Using the QUADAS-2 Tool

eFigure 3. Assessment of Funnel Plot Asymmetry and Contributed Heterogeneity for Studies Comparing Stress CMR With ICA

eFigure 4. Publication Bias Assessment

eFigure 5. Sensitivity Analysis for Diagnostic Accuracy of Stress CMR Compared With ICA

eFigure 6. Sensitivity Analysis for Diagnostic Accuracy of Stress CMR Compared With FFR

eTable 1. Details of Diagnostic Studies

eTable 2. Details of Prognostic Studies

eTable 3. Newcastle-Ottawa Scale for Quality Assessment of Prognostic Studies

eTable 4. Subgroup Analyses for Diagnostic Test Accuracy Studies Comparing Stress CMR With ICA

eTable 5. Subgroup Analyses for Diagnostic Test Accuracy Studies Comparing Stress CMR With FFR

eReferences

Data Sharing Statement

See More About

Select your interests.

Customize your JAMA Network experience by selecting one or more topics from the list below.

• Academic Medicine
• Acid Base, Electrolytes, Fluids
• Allergy and Clinical Immunology
• American Indian or Alaska Natives
• Anesthesiology
• Anticoagulation
• Art and Images in Psychiatry
• Artificial Intelligence
• Assisted Reproduction
• Bleeding and Transfusion
• Caring for the Critically Ill Patient
• Challenges in Clinical Electrocardiography
• Climate and Health
• Climate Change
• Clinical Challenge
• Clinical Decision Support
• Clinical Implications of Basic Neuroscience
• Clinical Pharmacy and Pharmacology
• Complementary and Alternative Medicine
• Consensus Statements
• Coronavirus (COVID-19)
• Critical Care Medicine
• Cultural Competency
• Dental Medicine
• Dermatology
• Diabetes and Endocrinology
• Diagnostic Test Interpretation
• Drug Development
• Electronic Health Records
• Emergency Medicine
• End of Life, Hospice, Palliative Care
• Environmental Health
• Equity, Diversity, and Inclusion
• Facial Plastic Surgery
• Gastroenterology and Hepatology
• Genetics and Genomics
• Genomics and Precision Health
• Global Health
• Guide to Statistics and Methods
• Hair Disorders
• Health Care Delivery Models
• Health Care Economics, Insurance, Payment
• Health Care Quality
• Health Care Reform
• Health Care Safety
• Health Care Workforce
• Health Disparities
• Health Inequities
• Health Policy
• Health Systems Science
• History of Medicine
• Hypertension
• Images in Neurology
• Implementation Science
• Infectious Diseases
• Innovations in Health Care Delivery
• JAMA Infographic
• Law and Medicine
• Leading Change
• Less is More
• LGBTQIA Medicine
• Lifestyle Behaviors
• Medical Coding
• Medical Devices and Equipment
• Medical Education
• Medical Education and Training
• Medical Journals and Publishing
• Mobile Health and Telemedicine
• Narrative Medicine
• Neuroscience and Psychiatry
• Notable Notes
• Nutrition, Obesity, Exercise
• Obstetrics and Gynecology
• Occupational Health
• Ophthalmology
• Orthopedics
• Otolaryngology
• Pain Medicine
• Palliative Care
• Pathology and Laboratory Medicine
• Patient Care
• Patient Information
• Performance Improvement
• Performance Measures
• Perioperative Care and Consultation
• Pharmacoeconomics
• Pharmacoepidemiology
• Pharmacogenetics
• Pharmacy and Clinical Pharmacology
• Physical Medicine and Rehabilitation
• Physical Therapy
• Physician Leadership
• Population Health
• Primary Care
• Professional Well-being
• Professionalism
• Psychiatry and Behavioral Health
• Public Health
• Pulmonary Medicine
• Regulatory Agencies
• Reproductive Health
• Research, Methods, Statistics
• Resuscitation
• Rheumatology
• Risk Management
• Scientific Discovery and the Future of Medicine
• Shared Decision Making and Communication
• Sleep Medicine
• Sports Medicine
• Stem Cell Transplantation
• Substance Use and Addiction Medicine
• Surgical Innovation
• Surgical Pearls
• Teachable Moment
• Technology and Finance
• The Art of JAMA
• The Arts and Medicine
• The Rational Clinical Examination
• Tobacco and e-Cigarettes
• Translational Medicine
• Trauma and Injury
• Treatment Adherence
• Ultrasonography
• Users' Guide to the Medical Literature
• Vaccination
• Venous Thromboembolism
• Veterans Health
• Women's Health
• Workflow and Process
• Wound Care, Infection, Healing

Others Also Liked

• Download PDF
• X Facebook More LinkedIn

Ricci F , Khanji MY , Bisaccia G, et al. Diagnostic and Prognostic Value of Stress Cardiovascular Magnetic Resonance Imaging in Patients With Known or Suspected Coronary Artery Disease : A Systematic Review and Meta-analysis . JAMA Cardiol. 2023;8(7):662–673. doi:10.1001/jamacardio.2023.1290

Manage citations:

© 2024

• Permissions

Diagnostic and Prognostic Value of Stress Cardiovascular Magnetic Resonance Imaging in Patients With Known or Suspected Coronary Artery Disease : A Systematic Review and Meta-analysis

• 1 Department of Neuroscience, Imaging and Clinical Sciences, Gabriele d’Annunzio University of Chieti-Pescara, Chieti, Italy
• 2 Department of Clinical Sciences, Lund University, Malmö, Sweden
• 3 William Harvey Research Institute, Barts Biomedical Research Centre, National Institute for Health and Care Research, Queen Mary University London, Charterhouse Square, London, United Kingdom
• 4 Newham University Hospital, Barts Health NHS Trust, London, United Kingdom
• 5 Barts Heart Centre, St Bartholomew’s Hospital, Barts Health NHS Trust, West Smithfield, London, United Kingdom
• 6 Department of Cardiac, Thoracic and Vascular Sciences and Public Health, University of Padova, Padova, Italy
• 7 Cardiology Unit, Rimini Hospital, Local Health Authority of Romagna, Rimini, Italy
• 8 Department of Cardiology, Karolinska University Hospital, Stockholm, Sweden
• 9 Department of Medicine, Karolinska Institute, Stockholm, Sweden
• 10 The Alan Turing Institute, London, United Kingdom
• 11 Health Data Research UK, London, United Kingdom
• 12 Royal Brompton and Harefield Hospitals, Guys and St Thomas NHS Trust London, London, United Kingdom
• 13 School of Biomedical Engineering and Imaging Sciences, Faculty of Life Sciences and Medicine, Kings College London, London, United Kingdom

Question   What is the diagnostic and prognostic value of stress cardiovascular magnetic resonance imaging (CMR) for the evaluation of stable chest pain?

Findings   In this systematic review and meta-analysis pooling 74 470 patients with stable chest pain over 381 357 person-years of follow-up, stress CMR yielded high diagnostic accuracy and accurate risk stratification in patients with known or suspected coronary artery disease, particularly when 3-T imaging was used. The presence of stress-inducible ischemia and late gadolinium enhancement was associated with higher mortality and likelihood of cardiovascular events, while normal stress CMR results were associated with a lower likelihood of cardiovascular events for at least 3.5 years.

Meaning   These findings suggest that combined assessment of inducible myocardial ischemia and late gadolinium enhancement by stress CMR is an accurate method to diagnose and risk stratify patients with stable chest pain and known or suspected coronary artery disease.

Importance   The clinical utility of stress cardiovascular magnetic resonance imaging (CMR) in stable chest pain is still debated, and the low-risk period for adverse cardiovascular (CV) events after a negative test result is unknown.

Objective   To provide contemporary quantitative data synthesis of the diagnostic accuracy and prognostic value of stress CMR in stable chest pain.

Data Sources   PubMed and Embase databases, the Cochrane Database of Systematic Reviews, PROSPERO, and the ClinicalTrials.gov registry were searched for potentially relevant articles from January 1, 2000, through December 31, 2021.

Study Selection   Selected studies evaluated CMR and reported estimates of diagnostic accuracy and/or raw data of adverse CV events for participants with either positive or negative stress CMR results. Prespecified combinations of keywords related to the diagnostic accuracy and prognostic value of stress CMR were used. A total of 3144 records were evaluated for title and abstract; of those, 235 articles were included in the full-text assessment of eligibility. After exclusions, 64 studies (74 470 total patients) published from October 29, 2002, through October 19, 2021, were included.

Data Extraction and Synthesis   This systematic review and meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Main Outcomes and Measures   Diagnostic odds ratios (DORs), sensitivity, specificity, area under the receiver operating characteristic curve (AUROC), odds ratio (OR), and annualized event rate (AER) for all-cause death, CV death, and major adverse cardiovascular events (MACEs) defined as the composite of myocardial infarction and CV death.

Results   A total of 33 diagnostic studies pooling 7814 individuals and 31 prognostic studies pooling 67 080 individuals (mean [SD] follow-up, 3.5 [2.1] years; range, 0.9-8.8 years; 381 357 person-years) were identified. Stress CMR yielded a DOR of 26.4 (95% CI, 10.6-65.9), a sensitivity of 81% (95% CI, 68%-89%), a specificity of 86% (95% CI, 75%-93%), and an AUROC of 0.84 (95% CI, 0.77-0.89) for the detection of functionally obstructive coronary artery disease. In the subgroup analysis, stress CMR yielded higher diagnostic accuracy in the setting of suspected coronary artery disease (DOR, 53.4; 95% CI, 27.7-103.0) or when using 3-T imaging (DOR, 33.2; 95% CI, 19.9-55.4). The presence of stress-inducible ischemia was associated with higher all-cause mortality (OR, 1.97; 95% CI, 1.69-2.31), CV mortality (OR, 6.40; 95% CI, 4.48-9.14), and MACEs (OR, 5.33; 95% CI, 4.04-7.04). The presence of late gadolinium enhancement (LGE) was associated with higher all-cause mortality (OR, 2.22; 95% CI, 1.99-2.47), CV mortality (OR, 6.03; 95% CI, 2.76-13.13), and increased risk of MACEs (OR, 5.42; 95% CI, 3.42-8.60). After a negative test result, pooled AERs for CV death were less than 1.0%.

Conclusion and Relevance   In this study, stress CMR yielded high diagnostic accuracy and delivered robust prognostication, particularly when 3-T scanners were used. While inducible myocardial ischemia and LGE were associated with higher mortality and risk of MACEs, normal stress CMR results were associated with a lower risk of MACEs for at least 3.5 years.

Coronary artery disease (CAD) is the leading cause of cardiovascular (CV) morbidity and mortality worldwide. Noninvasive imaging plays a central role in the 2019 European Society of Cardiology guidelines on chronic coronary syndromes 1 and the 2021 American Heart Association/American College of Cardiology (AHA/ACC) guidelines on chest pain. 2 Evaluation of stress-inducible myocardial ischemia by assessment of perfusion reserve or regional wall motion abnormalities is a key element in the diagnostic workup of patients with stable chest pain and intermediate to high pretest probability of CAD. 1 , 3

New recommendations for the use of noninvasive imaging in coronary syndromes developed by a transatlantic intersociety task force 4 endorse the use of stress cardiovascular magnetic resonance imaging (CMR) to detect ischemia and guide clinical decision-making in patients with a high intermediate pretest clinical likelihood of CAD. Consistent with this endorsement, the 2021 AHA/ACC guidelines for the evaluation and diagnosis of chest pain delivered class 1 and 2A recommendations for stress CMR as first-line functional investigation for the evaluation of chest pain in patients with known or suspected CAD who are at intermediate risk. 2

Coronary artery disease is one of the primary indications for CMR, 5 , 6 and the use of stress CMR has been steadily growing worldwide. 6 However, contemporary data on the diagnostic accuracy and prognostic value of stress CMR in patients with known or suspected CAD are currently lacking. After 20 years of clinical use and the completion of large multicenter observational studies 7 , 8 and randomized clinical trials, 9 , 10 which were not included in previous systematic reviews and meta-analyses, 11 - 14 we appraised the best available contemporary evidence to deliver an updated quantitative synthesis on the diagnostic accuracy and prognostic value of stress CMR for the assessment of chest pain.

This systematic review and meta-analysis was planned, conducted, and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses ( PRISMA ) guideline for design, analysis, and reporting of meta-analyses of randomized and observational studies 15 and the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. 16 A review protocol was prospectively registered on PROSPERO ( CRD42022299275 ).

We searched PubMed and Embase databases, the Cochrane Database of Systematic Reviews, the PROSPERO database, and the ClinicalTrials.gov registry for potentially relevant articles from January 1, 2000, through December 31, 2021 (eFigure 1 in Supplement 1 ). We used 2 prespecified combinations of keywords related to the diagnostic accuracy of stress CMR (eMethods in Supplement 1 ). We also searched reference lists of all identified articles for additional relevant studies, including hand searching reviews and published meta-analyses.

Two authors (G.B. and A.D.C.) performed the screening of titles and abstracts, reviewed full-text articles, and determined their eligibility. Discrepancies were resolved by consensus with other reviewers (F.R., M.Y.K., and A.C.). The review process was not blinded to study results. Studies were eligible if they met the following criteria: (1) were published as a full-length article, (2) were written in the English language, (3) had a prospective or retrospective study design, (4) enrolled 100 or more patients aged 18 years or older, and (5) reported estimates of the diagnostic accuracy of stress CMR compared with invasive coronary angiography (ICA) or fractional flow reserve (FFR) as the reference test and/or reported raw data for all-cause death, CV death, and major adverse cardiovascular events (MACEs, defined as the composite of CV death and myocardial infarction [MI]) for study participants with either positive or negative stress CMR scans. Studies were eligible regardless of whether they included patients who were referred for suspected or known CAD and regardless of the technique used for evaluation of inducible ischemia (ie, wall motion analysis or perfusion; qualitative, semiquantitative, or fully quantitative).

Two investigators (G.B. and A.D.C.) abstracted relevant data of patient populations, study-level characteristics, and outcomes from original eligible sources. The ascertainment of clinical events was accepted as reported. The quality of eligible studies was evaluated using the Quality Assessment of Diagnostic Accuracy Studies, version 2 (QUADAS-2) tool 17 for diagnostic studies and the Newcastle-Ottawa Scale 18 for prognostic studies.

Categorical variables were reported as percentages and continuous variables as means with SDs or medians with IQRs, as appropriate. We used the inverse variance heterogeneity model for the meta-analysis of diagnostic studies, which proved superior to the standard bivariate model. 19 For each study, raw data of true-positive, true-negative, false-positive, and false-negative results were either extracted from the study or generated from reported diagnostic estimates. Diagnostic odds ratios (DORs), area under the receiver operating characteristic curve (AUROC), sensitivity, specificity, negative likelihood ratio (NLR), and positive likelihood ratio (PLR) were calculated. An ROC plot was used to summarize study-level findings. Pooled estimates of sensitivity and specificity for stress CMR derived from the meta-analysis were used to generate a leaf plot illustrating the association between pretest and posttest probability of CAD.

In the prognostic meta-analysis, summary effect sizes for all-cause death, CV death, and MI were calculated primarily for the presence or absence of inducible ischemia in addition to late gadolinium enhancement (LGE). A random-effects model was used, and study-specific odds ratios (ORs) were pooled using the Mantel-Haenszel method for each study outcome. The Hartung-Knapp adjustment 20 was applied to all analyses except for those with 3 or fewer studies per group. Mean effects were not calculated for outcomes reported by fewer than 3 studies. Interstudy heterogeneity was assessed using the I 2 statistic and represented as a Baujat plot. 21 Significant heterogeneity was defined as I 2 values of 50% or greater. The z statistic was computed for each end point of interest, and the results were considered statistically significant at 1-sided P  < .05.

Meta-analysis results were presented using classic forest plots with point estimates of the effect size and 95% CIs, with square area indicating study weight. A jackknife sensitivity analysis was performed for each outcome to evaluate the robustness of the results and the effect of each study on the summary estimate of effect. The likelihood of publication bias was assessed using funnel plots by displaying individual study ORs with 95% CIs for the end points of interest, with the addition of the nonparametric trim-and-fill procedure to adjust for funnel plot asymmetry by generating hypothetical missing studies; for all models including more than 10 studies, funnel plot asymmetry was also evaluated using tests proposed by Deeks et al 22 for diagnostic studies and Egger et al 23 for prognostic studies (with 1-sided P  < .10 indicating significant publication bias).

Subgroup analyses were performed to investigate possible sources of heterogeneity and to assess the effect of selected variables, including sample size, sex, CAD prevalence, thresholds of diameter stenosis, year of publication, magnetic field strength, and stressor agent. Annualized event rates (AERs) for studies were calculated by dividing the number of events by the follow-up duration. The low-risk period was defined as the mean interval during which the patient group with a negative test remained lower than the threshold of 1% of the cumulative MACE rate. 24 All statistical analyses were performed using R software, version 4.1.0. (R Foundation for Statistical Computing) (R packages and functions are detailed in the eMethods in Supplement 1 ).

Of 3144 records (1152 diagnostic studies and 1992 prognostic studies) identified and retrieved for title and abstract evaluation, 2909 (1038 diagnostic studies and 1871 prognostic studies) were excluded, resulting in 235 potentially relevant articles (114 diagnostic studies and 121 prognostic studies) included in the full-text assessment of eligibility. After exclusions, 64 studies, 7 - 10 , 25 - 83 including 33 diagnostic studies 8 , 10 , 25 - 55 pooling 7814 individuals and 31 prognostic studies 7 , 9 , 25 , 56 - 83 pooling 67 080 individuals, published between October 29, 2002, and October 19, 2021, were included in the meta-analysis (eFigure 1 in Supplement 1 ). The study-level prevalence of CAD ranged between 11% and 83% in diagnostic studies. The mean (SD) follow-up was 3.5 (2.1) years (range, 0.9-8.8 years), for a total of 381 357 person-years among 74 470 total patients. The overall quality of included studies was high (eFigure 2 and eTable 3 in Supplement 1 ). The main characteristics of studies included in the diagnostic 8 , 10 , 25 - 55 and prognostic 7 , 9 , 25 , 56 - 83 meta-analyses are summarized in eTables 1 and 2 in Supplement 1 .

The diagnostic accuracy of stress CMR compared with ICA as the reference test was reported in 30 studies 8 , 10 , 25 - 52 pooling 7496 symptomatic patients, of whom 537 had known CAD and 2825 had suspected CAD. In the per-patient analysis, stress CMR yielded a pooled DOR of 19.1 (95% CI, 12.6-29.1), a sensitivity of 84% (95% CI, 79%-88%), a specificity of 79% (95% CI, 73%-84%), a PLR of 4.0 (95% CI, 3.0-5.3), an NLR of 0.2 (95% CI, 0.2-0.3), and an AUROC of 0.81 (95% CI, 0.78-0.84) for the detection of anatomically obstructive CAD ( Figure 1 A). In the per-vessel analysis, stress CMR yielded a pooled DOR of 21.0 (95% CI, 10.2-43.4), a sensitivity of 72% (95% CI, 61%-81%), a specificity of 89% (95% CI, 82%-94%), a PLR of 6.7 (95% CI, 3.8-11.8), an NLR of 0.3 (95% CI, 0.2-0.5), and an AUROC of 0.82 (95% CI, 0.76-0.87).

The diagnostic accuracy of stress CMR compared with invasive FFR as the reference test was reported in 8 studies 10 , 27 , 37 , 44 , 45 , 53 - 55 pooling 1196 symptomatic patients, of whom 354 had known CAD and 593 had suspected CAD. In the per-patient analysis, stress CMR yielded a pooled DOR of 26.4 (95% CI, 10.6-65.9), a sensitivity of 81% (95% CI, 68%-89%), a specificity of 86% (95% CI, 75%-93%), a PLR of 5.8 (95% CI, 3.0-11.4), an NLR of 0.2 (95% CI, 0.1-0.4), and an AUROC of 0.84 (95% CI, 0.77-0.89) for the detection of functionally obstructive CAD ( Figure 1 B). In the per-vessel analysis, stress CMR yielded a pooled DOR of 24.1 (95% CI, 5.5-105.4), a sensitivity of 70% (95% CI, 46%-86%), a specificity of 91% (95% CI, 74%-97%), a PLR of 8.0 (95% CI, 2.4-26.5), an NLR of 0.3 (95% CI, 0.1-0.8), and an AUROC of 0.83 (95% CI, 0.70-0.91).

A total of 11 studies 9 , 56 - 65 pooling 51 166 individuals reported all-cause mortality. The presence of inducible ischemia was associated with a 2-fold increased mortality (OR, 1.97; 95% CI, 1.69-2.31; P  = .002) ( Figure 2 A). The presence of LGE was associated with 2-fold increased mortality (OR, 2.22; 95% CI, 1.99-2.47; P  < .001) ( Figure 3 A). Pooled AERs for all-cause mortality were 2.97% in patients with inducible ischemia vs 1.40% in patients without inducible ischemia ( P  < .001) and 4.46% in patients with LGE vs 2.30% in patients without LGE ( P  < .001) ( Figure 4 A).

A total of 14 studies 9 , 62 , 65 - 76 pooling 12 252 individuals reported CV death data. The presence of inducible ischemia detected by stress CMR was associated with 6-fold increased CV mortality (OR, 6.40; 95% CI, 4.48-9.14; P  < .001) ( Figure 2 B). The presence of LGE was associated with 6-fold increased CV mortality (OR, 6.03; 95% CI, 2.76-13.13; P  < .001) ( Figure 3 B). Pooled AERs for CV death were 2.51% in patients with inducible ischemia vs 0.59% in patients without inducible ischemia ( P  < .001) and 2.51% in patients with LGE vs 0.71% in patients without LGE ( P  < .001) ( Figure 4 B).

A total of 22 studies 7 , 9 , 25 , 59 , 60 , 65 - 68 , 71 - 83 pooling 17 084 individuals reported MACE data. The presence of inducible ischemia was associated with a 5-fold higher likelihood of incident MACEs (OR, 5.33; 95% CI, 4.04-7.04; P  < .001) ( Figure 5 ). The presence of LGE was associated with a 5-fold higher likelihood of MACEs (OR, 5.42; 95% CI, 3.42-8.60; P  < .001) ( Figure 3 C). Pooled AERs for MACEs were 4.31% in patients with ischemia vs 0.98% in patients without ischemia ( P  < .001)and 2.90% in patients with LGE vs 0.78% in patients without LGE ( P  < .001) ( Figure 4 A).

Combining ischemia and LGE information, we documented the highest AER when both were present and the lowest AER when both were absent ( Figure 4 B). At a mean follow-up of 3.5 years, normal stress CMR results featuring the absence of both inducible ischemia and LGE were associated with a pooled AER of 0.58%, while the presence of both ischemia and LGE yielded a pooled AER of 4.24% ( P  < .001).

According to the QUADAS-2 tool, risk of bias was low in 29 27 - 55 of 33 diagnostic studies 8 , 10 , 25 - 55 (eFigure 2 in Supplement 1 ). Of 31 prognostic studies, 7 , 9 , 25 , 56 - 83 15 studies 7 , 9 , 25 , 56 , 60 , 62 - 64 , 66 , 68 , 73 , 74 , 76 , 78 , 80 scored 9 stars and 16 studies 57 - 59 , 61 , 65 , 67 , 69 - 72 , 75 , 77 , 79 , 81 - 83 scored 8 stars according to the Newcastle-Ottawa Scale (eTable 3 in Supplement 1 ). In ICA studies, 8 , 10 , 25 - 52 the Deeks test 22 ruled out small study bias and publication bias ( P  = .34) (eFigure 3 in Supplement 1 ). The Deeks test was not performed in FFR studies 10 , 27 , 37 , 44 , 45 , 53 - 55 because the number of studies was insufficient. With regard to prognostic studies, 7 , 9 , 25 , 56 - 83 we ruled out publication bias by visual inspection of funnel plots and the Egger test 23 of intercept, which was nonsignificant for each outcome (eFigure 4 in Supplement 1 ).

Results of the subgroup analysis are summarized in eTables 4 and 5 in Supplement 1 . Stress CMR yielded higher diagnostic performance for the detection of anatomically and functionally obstructive CAD in 2 scenarios: suspected CAD (DOR, 53.4; 95% CI, 27.7-103.0) and 3-T imaging (DOR, 33.2; 95% CI, 19.9-55.4). In FFR studies, 10 , 27 , 37 , 44 , 45 , 53 - 55 higher diagnostic accuracy was observed when women were assessed or when the FFR cutoff was lowered to 0.75. In ICA studies, 8 , 10 , 25 - 52 quantitative assessment yielded higher DORs and specificity compared with visual assessment, and dipyridamole achieved higher accuracy overall compared with adenosine.

Two diagnostic studies 10 , 51 were visually and quantitatively identified as outliers in the ICA analysis 8 , 10 , 25 - 52 (eFigure 3 in Supplement 1 ). Removal of the 2 outliers 10 , 51 increased diagnostic accuracy, with a pooled DOR of 25.2 (eFigure 5 in Supplement 1 ). In the FFR analysis, 10 , 27 , 37 , 44 , 45 , 53 - 55 removal of the single outlier 10 improved diagnostic summary estimates, attaining a pooled DOR of 41.3 (eFigure 6 in Supplement 1 ). No single prognostic study changed the pooled OR for each end point of interest.

The current systematic review and meta-analysis covered the last 20 years of clinical research in the field of stress CMR using state-of-the-art statistical methods for quantitative data synthesis. We provided the largest summary evidence available by pooling 74 470 patients and 381 357 person-years of follow-up. Our findings reaffirmed that stress CMR yields high diagnostic accuracy, robust cardiac prognostication, and accurate risk stratification in patients with stable chest pain and known or suspected CAD. Our analysis was focused on symptomatic patients, consistent with current international guideline indications on deferring or eliminating unnecessary testing when the diagnostic yield is low or when individuals are asymptomatic. 1 , 2

Stress CMR consistently delivered high diagnostic accuracy across multiple clinical scenarios and temporal pattern analyses. This accuracy was particularly evident with regard to the detection of functionally obstructive lesions assessed by FFR, which was found to provide optimum balance between myocardial revascularization and medical treatment in the FAME (Fractional Flow Reserve Versus Angiography in Multivessel Evaluation) trials. 84 , 85 In addition to results from previous meta-analyses, 86 , 87 our findings provide evidence that stress CMR has better diagnostic performance in the setting of suspected CAD or when using 3-T imaging due to improved contrast resolution 88 - 90 and quantitative perfusion assessment, which can be advantageous to better identify the extent of disease or peri-infarction ischemia in multivessel CAD compared with visual assessment alone and can more accurately detect microvascular disease and the effectiveness of the stressor agents. 91 The signal of dipyridamole outperforming adenosine was intriguing and possibly reflected the incremental diagnostic value of combined perfusion and wall motion assessment. 75 This finding requires careful interpretation and prospective verification in regadenoson studies and needs to be weighed against the cost, potential tolerability, and benefit of the stressor agents. 92

In our diagnostic meta-analysis, 2 studies 10 , 51 were identified as outliers that had a diagnostic yield lower than the mean for stress CMR. The Dan-NICAD (Danish Study of Non-Invasive Diagnostic Testing in Coronary Artery Disease) randomized clinical trial 10 enrolled patients with a low to intermediate pretest probability of CAD and an abnormal coronary computed tomography (CT) angiographic scan before CMR and found low sensitivity for second-line perfusion investigations. However, the specific study design could have led to selection bias and potentially impacted diagnostic estimates. 93 The MR-IMPACT II (Magnetic Resonance Imaging for Myocardial Perfusion Assessment in Coronary Artery Disease Trial II) study 51 compared stress CMR with single-photon emission CT in a population with intermediate CAD prevalence (49%). However, this study 51 had a fairly high number of patients with previous MI (27%); in these patients, it can be more difficult to discriminate myocardial scarring and residual ischemia, and the expected higher prevalence of microvascular disease in this population can inflate the number of false-positive findings. This multicenter study 51 enrolling participants from 33 different institutions aimed to frame a realistic clinical environment not restricted to high-volume leading centers. In both studies, 10 , 51 measurements were performed by an independent core laboratory with readers fully blinded to additional patient information and results, limiting the bias of the clinical context when reporting stress CMR studies.

When interpreting these findings, we should remember that myocardial ischemia exists on a continuum, and binary categorizations have inherent limitations. Furthermore, shortcomings in the accuracy of established invasive gold standards must be carefully considered. Notably, FFR was first calibrated against noninvasive tests, 94 including bicycle exercise testing, thallium scintigraphy, and stress echocardiography with dobutamine, which were themselves validated against ICA as the reference test, resulting in problematic circular thinking. 95 , 96 An FFR threshold of 0.80 or lower has been adopted into clinical practice guidelines 97 as an actionable value to guide revascularization despite robust evidence supporting larger treatment benefit at lower FFR values 98 , 99 ; our findings indicate better agreement at an FFR threshold of 0.75.

The 2019 MR-INFORM (MR Perfusion Imaging to Guide Management of Patients With Stable Coronary Artery Disease) trial 9 randomized 918 symptomatic patients at high pretest probability of CAD to undergo ICA plus FFR vs stress CMR-guided assessment. The MACE rate and percentage of patients free of angina were similar for both strategies at 1 year; however, the use of stress CMR was associated with a lower incidence of downstream ICA and coronary revascularization than the use of FFR. 9 Similar findings have been reported in the setting of low-risk acute coronary syndromes by a network meta-analysis of diagnostic randomized clinical trials, 100 which found that stress CMR was associated with fewer referrals to downstream ICA than coronary CT angiography or other noninvasive imaging modalities and without obvious consequences for the subsequent risk of MI.

This evidence translates into a distinctly favorable cost-effectiveness profile for stress CMR compared with its relevant comparators. 101 According to a cost-effectiveness analysis 102 comparing different first-line diagnostic approaches for stable chest pain and a decision-analytic model estimating lifetime health care costs and quality-adjusted life-years derived from the multicenter SPINS (Stress CMR Perfusion Imaging in the United States) study, 103 stress CMR strongly dominated single-photon emission CT and coronary CT angiography strategies when considering either all MACEs or CV mortality alone. Thus, having access to CMR is a beneficial situation for patients and may lead to substantial cost savings by reducing the need for additional unnecessary tests and revascularization procedures. 104 , 105

The prognostic value of noninvasive cardiac assessments was the objective of a previous meta-analysis 13 raising the possibility of clinical equipoise for estimation of CV death and MI. While the message that any negative test conveys excellent prognosis may be reassuring to patients and challenges the need for further downstream testing, the posttest probability of disease needs to be adjusted for baseline population event risk and should always be carefully interpreted in the context of pretest probability and prevalence of disease and according to the clinical scenario. In our analysis, the presence of inducible ischemia on stress CMR was a robust estimator of increased mortality and MACEs, further heightened by the presence of LGE. Conversely, normal stress CMR results were associated with a low incidence of MACEs, yielding a low-risk posttest period of at least 3.5 years. Our data were consistent with the results of previous meta-analyses 106 , 107 and findings of the European Cardiovascular Magnetic Resonance (EuroCMR) registry, 5 in which patients with suspected CAD and a negative stress CMR result experienced an AER for aborted sudden cardiac death and nonfatal MI of less than 1%.

Ultimately, the prognostic value of stress CMR, either performed with vasodilators or dobutamine, provides incremental risk stratification in patients with stable chest pain. 65 , 80 Further studies are needed to establish the optimal CMR method for absolute quantification of myocardial blood flow and the optimal ischemic threshold associated with larger treatment effect, which would be useful to identify patients who would most benefit from myocardial revascularization vs safe deferral.

This study has several strengths. We summarized the largest evidence available, making use of the best methods for quantitative synthesis, and we provide robust estimates of the diagnostic and prognostic value of stress CMR. We provide new information on the duration of the low-risk period for MACEs after a normal stress CMR result. This knowledge has the potential to inform future clinical guidelines about ideal intervals for repeat imaging and to provide useful guidance for subsequent management of assessment strategies among symptomatic patients with initial normal imaging results or subclinical disease. 108 Results of the subgroup analyses also suggest better diagnostic performance of stress CMR in the setting of suspected CAD, especially when using 3-T imaging and fully quantitative approaches.

This study also has limitations. First, we did not compare the yield of stress CMR with other imaging modalities because it was beyond the scope of the current work, and literature specifically addressing these topics already exists. 109 - 111 Second, our results are mostly derived from observational studies reflecting different guideline recommendations across 2 decades of practice. Within this time span, thresholds for coronary stenosis have changed, 112 methods for estimation of pretest probabilities of obstructive CAD have been updated and recalibrated, 1 , 2 and CMR protocols have been implemented, including quantitative perfusion assessment, 61 new tools for evaluation of stress adequacy, 113 - 115 more widespread use of regadenoson, 116 and other disruptive technical innovations. 117 - 119 Third, we recognize there is a lack of information about medical therapy, completeness of myocardial revascularization, extent of inducible ischemia, degree of myocardial fibrosis, and prevalence of microvascular dysfunction. Despite the intrinsic challenges and limitations of study-level meta-analysis, including limited adjustment for confounding factors and susceptibility to the ecological fallacy, we attempted to synthesize the results in a robust manner that addressed potential bias.

This systematic review and meta-analysis found that in patients with stable chest pain and known or suspected CAD, stress CMR yielded high diagnostic accuracy to detect both anatomically and functionally significant CAD, with 3-T and quantitative perfusion approaches delivering higher diagnostic performance. Stress CMR also provided robust prognostic information and accurate risk stratification. While the presence of ischemia and LGE were associated with higher CV risk and mortality, normal stress CMR results were associated with a lower likelihood of MACEs for at least 3.5 years. These findings suggest that combined assessment of inducible myocardial ischemia and LGE by stress CMR is an accurate method to diagnose and risk stratify patients with known or suspected CAD.

Accepted for Publication: April 12, 2023.

Published Online: June 7, 2023. doi:10.1001/jamacardio.2023.1290

Corresponding Author: Chiara Bucciarelli-Ducci, MD, PhD, Royal Brompton Hospital, Sydney Street, London SW3 6NP, United Kingdom ( [email protected] ).

Author Contributions: Drs Ricci and Bisaccia had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Ricci, Khanji, Bisaccia, Di Cesare, Mantini, Zimarino, Fedorowski, Gallina, Petersen, Bucciarelli-Ducci.

Acquisition, analysis, or interpretation of data: Ricci, Khanji, Bisaccia, Cipriani, Di Cesare, Ceriello, Bucciarelli-Ducci.

Drafting of the manuscript: Ricci, Bisaccia, Di Cesare.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Ricci, Bisaccia.

Administrative, technical, or material support: Khanji, Cipriani, Di Cesare.

Supervision: Khanji, Cipriani, Di Cesare, Ceriello, Mantini, Zimarino, Fedorowski, Gallina, Petersen, Bucciarelli-Ducci.

Conflict of Interest Disclosures: Prof Fedorowski reported receiving personal fees from argenx, Finapres Medical Systems, and Medtronic outside the submitted work. Prof Petersen reported receiving personal fees from Circle Cardiovascular Imaging outside the submitted work. Prof Bucciarelli-Ducci reported receiving personal fees from Bayer, Circle Cardiovascular Imaging, Siemens Healthineers, and the Society for Cardiovascular Magnetic Resonance (for which she serves as chief executive officer) outside the submitted work. No other disclosures were reported.

Funding/Support: This study was supported by the 2020 Search for Excellence Starting Grant, Gabriele d’Annunzio University of Chieti-Pescara, Italy.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2 .

• Register for email alerts with links to free full-text articles
• Access PDFs of free articles
• Manage your interests
• Save searches and receive search alerts

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Published: 26 August 2024

Predicting treatment outcomes in major depressive disorder using brain magnetic resonance imaging: a meta-analysis

• Fenghua Long   ORCID: orcid.org/0000-0002-8287-9009 1 , 2   na1 ,
• Yufei Chen 1 , 2   na1 ,
• Qian Zhang 1 , 2   na1 ,
• Qian Li 1 , 2 ,
• Yaxuan Wang 1 , 2 ,
• Yitian Wang 1 , 2 ,
• Haoran Li 1 , 2 ,
• Youjin Zhao 1 , 2 ,
• Robert K. McNamara 3 ,
• Melissa P. DelBello 3 ,
• John A. Sweeney 1 , 3 ,
• Qiyong Gong   ORCID: orcid.org/0000-0002-5912-4871 1 , 2 &
• Fei Li   ORCID: orcid.org/0000-0002-4737-5710 1 , 2  

Molecular Psychiatry ( 2024 ) Cite this article

59 Accesses

Metrics details

• Neuroscience
• Prognostic markers

Recent studies have provided promising evidence that neuroimaging data can predict treatment outcomes for patients with major depressive disorder (MDD). As most of these studies had small sample sizes, a meta-analysis is warranted to identify the most robust findings and imaging modalities, and to compare predictive outcomes obtained in magnetic resonance imaging (MRI) and studies using clinical and demographic features. We conducted a literature search from database inception to July 22, 2023, to identify studies using pretreatment clinical or brain MRI features to predict treatment outcomes in patients with MDD. Two meta-analyses were conducted on clinical and MRI studies, respectively. The meta-regression was employed to explore the effects of covariates and compare the predictive performance between clinical and MRI groups, as well as across MRI modalities and intervention subgroups. Meta-analysis of 13 clinical studies yielded an area under the curve (AUC) of 0.73, while in 44 MRI studies, the AUC was 0.89. MRI studies showed a higher sensitivity than clinical studies (0.78 vs. 0.62, Z = 3.42, P  = 0.001). In MRI studies, resting-state functional MRI (rsfMRI) exhibited a higher specificity than task-based fMRI (tbfMRI) (0.79 vs. 0.69, Z = −2.86, P  = 0.004). No significant differences in predictive performance were found between structural and functional MRI, nor between different interventions. Of note, predictive MRI features for treatment outcomes in studies using antidepressants were predominantly located in the limbic and default mode networks, while studies of electroconvulsive therapy (ECT) were restricted mainly to the limbic network. Our findings suggest a promise for pretreatment brain MRI features to predict MDD treatment outcomes, outperforming clinical features. While tasks in tbfMRI studies differed, those studies overall had less predictive utility than rsfMRI data. Overlapping but distinct network-level measures predicted antidepressants and ECT outcomes. Future studies are needed to predict outcomes using multiple MRI features, and to clarify whether imaging features predict outcomes generally or differ depending on treatments.

This is a preview of subscription content, access via your institution

Access options

Subscribe to this journal

Receive 12 print issues and online access

251,40 € per year

only 20,95 € per issue

Buy this article

• Purchase on SpringerLink
• Instant access to full article PDF

Prices may be subject to local taxes which are calculated during checkout

Similar content being viewed by others

Individualized fMRI connectivity defines signatures of antidepressant and placebo responses in major depression

Breakthroughs and challenges for generating brain network-based biomarkers of treatment response in depression

Brain connectivity in major depressive disorder: a precision component of treatment modalities?

Data availability.

The data that support the findings of this study, along with the code utilized in the Methods section, are available from the corresponding authors upon reasonable request.

World Health Organization. Depression fact sheet. World Health Organization. Published December 2019. Accessed November 20, 2022. https://www.who.int/mediacentre/factsheets/fs369/en/ .

Chen X, Lu B, Li HX, Li XY, Wang YW, Castellanos FX, et al. The DIRECT consortium and the REST-meta-MDD project: towards neuroimaging biomarkers of major depressive disorder. Psychoradiology. 2022;2:32–42.

Article   PubMed   PubMed Central   Google Scholar  

Jain FA, Connolly CG, Reus VI, Meyerhoff DJ, Yang TT, Mellon SH, et al. Cortisol, moderated by age, is associated with antidepressant treatment outcome and memory improvement in Major Depressive Disorder: A retrospective analysis. Psychoneuroendocrinology. 2019;109:104386.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Sramek JJ, Murphy MF, Cutler NR. Sex differences in the psychopharmacological treatment of depression. Dialogues Clin Neurosci. 2016;18:447–57.

Riedel M, Möller HJ, Obermeier M, Adli M, Bauer M, Kronmüller K, et al. Clinical predictors of response and remission in inpatients with depressive syndromes. J Affect Disord. 2011;133:137–49.

Article   PubMed   Google Scholar  

Davis AK, Barrett FS, May DG, Cosimano MP, Sepeda ND, Johnson MW, et al. Effects of Psilocybin-Assisted Therapy on Major Depressive Disorder: A Randomized Clinical Trial. JAMA Psychiatry. 2021;78:481–89.

Zhou Y, Zhang Z, Wang C, Lan X, Li W, Zhang M, et al. Predictors of 4-week antidepressant outcome in patients with first-episode major depressive disorder: An ROC curve analysis. J Affect Disord. 2022;304:59–65.

Article   PubMed   CAS   Google Scholar  

Lee Y, Ragguett RM, Mansur RB, Boutilier JJ, Rosenblat JD, Trevizol A, et al. Applications of machine learning algorithms to predict therapeutic outcomes in depression: A meta-analysis and systematic review. J Affect Disord. 2018;241:519–32.

Carstens L, Hartling C, Stippl A, Domke A-K, Herrera-Mendelez AL, Aust S, et al. A symptom-based approach in predicting ECT outcome in depressed patients employing MADRS single items. Eur Arc Psychiatry Clin Neurosci. 2021;271:1275–84.

Article   Google Scholar  

Krepel N, Rush AJ, Iseger TA, Sack AT, Arns M. Can psychological features predict antidepressant response to rTMS? A Discovery-Replication approach. Psychol Med. 2020;50:264–72.

Kautzky A, Dold M, Bartova L, Spies M, Vanicek T, Souery D, et al. Refining Prediction in Treatment-Resistant Depression: Results of Machine Learning Analyses in the TRD III Sample. J Clin Psychiatry. 2018;79:16m11385.

Li F, Sun H, Biswal BB, Sweeney JA, Gong Q. Artificial intelligence applications in psychoradiology. Psychoradiology. 2021;1:94–107.

Luo L, You W, DelBello MP, Gong Q, Li F. Recent advances in psychoradiology. Phys Med Biol. 2022;67:23TR01.

You W, Luo L, Yao L, Zhao Y, Li Q, Wang Y, et al. Impaired dynamic functional brain properties and their relationship to symptoms in never treated first-episode patients with schizophrenia. Schizophrenia (Heidelb). 2022;8:90.

Luo L, Li Q, Wang Y, He N, Wang Y, You W, et al. Shared and Disorder-Specific Alterations of Brain Temporal Dynamics in Obsessive-Compulsive Disorder and Schizophrenia. Schizophr Bull. 2023;49:1387–98.

Moreno-Ortega M, Prudic J, Rowny S, Patel GH, Kangarlu A, Lee S, et al. Resting state functional connectivity predictors of treatment response to electroconvulsive therapy in depression. Sci Rep. 2019;9:5071.

Godlewska BR, Browning M, Norbury R, Igoumenou A, Cowen PJ, Harmer CJ. Predicting Treatment Response in Depression: The Role of Anterior Cingulate Cortex. Int J Neuropsychopharmacol. 2018;21:988–96.

Hu X, Zhang L, Hu X, Lu L, Tang S, Li H, et al. Abnormal Hippocampal Subfields May Be Potential Predictors of Worse Early Response to Antidepressant Treatment in Drug-Naïve Patients With Major Depressive Disorder. J Magn Reson Imaging. 2019;49:1760–68.

Wu P, Zhang A, Sun N, Lei L, Liu P, Wang Y, et al. Cortical Thickness Predicts Response Following 2 Weeks of SSRI Regimen in First-Episode, Drug-Naive Major Depressive Disorder: An MRI Study. Front Psychiatry. 2021;12:751756.

Cohen SE, Zantvoord JB, Wezenberg BN, Bockting CLH, van Wingen GA. Magnetic resonance imaging for individual prediction of treatment response in major depressive disorder: a systematic review and meta-analysis. Transl Psychiatry. 2021;11:168.

Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. Ann Intern Med. 2009;151:W65–94.

Doebler P. mada: Meta-Analysis of Diagnostic Accuracy. R package version 0.5.11. https://r-forge.r-project.org/projects/mada/ . 2022.

Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw. 2010;33:1–22.

Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. Bmj. 2003;327:557–60.

Glas AS, Lijmer JG, Prins MH, Bonsel GJ, Bossuyt PM. The diagnostic odds ratio: a single indicator of test performance. J Clin Epidemiol. 2003;56:1129–35.

Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol. 2005;58:982–90.

Devillé WL, Buntinx F, Bouter LM, Montori VM, de Vet HC, van der Windt DA, et al. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC Med Res Methodol. 2002;2:9.

Jones CM, Athanasiou T. Summary receiver operating characteristic curve analysis techniques in the evaluation of diagnostic tests. Ann Thorac Surg. 2005;79:16–20.

Gowin JL, Manza P, Ramchandani VA, Volkow ND. Neuropsychosocial markers of binge drinking in young adults. Mol Psychiatry. 2021;26:4931–43.

Gosnell SN, Curtis KN, Velasquez K, Fowler JC, Madan A, Goodman W, et al. Habenular connectivity may predict treatment response in depressed psychiatric inpatients. J Affect Disord. 2019;242:211–19.

Jackson D, Turner R. Power analysis for random-effects meta-analysis. Res Synth Methods. 2017;8:290–302.

Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol. 2005;58:882–93.

Macaskill PTY, Deeks JJ, Gatsonis C. Chapter 9: Understanding meta-analysis. Draft version (4 October 2022) for inclusion in: Deeks JJ, Bossuyt PM, Leeflang MM, Takwoingi Y, editor(s). Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 2. London: Cochrane.

Leucht S, Fennema H, Engel RR, Kaspers-Janssen M, Szegedi A. Translating the HAM-D into the MADRS and vice versa with equipercentile linking. J Affect Disord. 2018;226:326–31.

Furukawa TA, Reijnders M, Kishimoto S, Sakata M, DeRubeis RJ, Dimidjian S, et al. Translating the BDI and BDI-II into the HAMD and vice versa with equipercentile linking. Epidemiol Psychiatr Sci. 2019 ;29:e24.

Takwoingi YDN, Schiller I, Rücker G, Jones HE, Partlett C, Macaskill P. Chapter 10: Undertaking meta-analysis. Draft version (4 October 2022) for inclusion in: Deeks JJ, Bossuyt PM, Leeflang MM, Takwoingi Y, editor(s). Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 2. London: Cochrane.

Hopman HJ, Chan SMS, Chu WCW, Lu H, Tse CY, Chau SWH, et al. Personalized prediction of transcranial magnetic stimulation clinical response in patients with treatment-refractory depression using neuroimaging biomarkers and machine learning. J Affect Disord. 2021;290:261–71.

Nakamura T, Tomita M, Horikawa N, Ishibashi M, Uematsu K, Hiraki T, et al. Functional connectivity between the amygdala and subgenual cingulate gyrus predicts the antidepressant effects of ketamine in patients with treatment-resistant depression. Neuropsychopharmacol Rep. 2021;41:168–78.

van Waarde JA, Scholte HS, van Oudheusden LJ, Verwey B, Denys D, van Wingen GA. A functional MRI marker may predict the outcome of electroconvulsive therapy in severe and treatment-resistant depression. Mol Psychiatry. 2015;20:609–14.

Goldstein-Piekarski AN, Staveland BR, Ball TM, Yesavage J, Korgaonkar MS, Williams LM. Intrinsic functional connectivity predicts remission on antidepressants: a randomized controlled trial to identify clinically applicable imaging biomarkers. Transl Psychiatry. 2018;8:57.

Braund TA, Breukelaar IA, Griffiths K, Tillman G, Palmer DM, Bryant R, et al. Intrinsic Functional Connectomes Characterize Neuroticism in Major Depressive Disorder and Predict Antidepressant Treatment Outcomes. Biol Psychiatry Cogn Neurosci Neuroimaging. 2021;7:276–84.

Korgaonkar MS, Goldstein-Piekarski AN, Fornito A, Williams LM. Intrinsic connectomes are a predictive biomarker of remission in major depressive disorder. Mol Psychiatry. 2020;25:1537–49.

Siegle GJ, Thompson WK, Collier A, Berman SR, Feldmiller J, Thase ME, et al. Toward clinically useful neuroimaging in depression treatment: prognostic utility of subgenual cingulate activity for determining depression outcome in cognitive therapy across studies, scanners, and patient characteristics. Arch Gen Psychiatry. 2012;69:913–24.

Crane NA, Jenkins LM, Bhaumik R, Dion C, Gowins JR, Mickey BJ, et al. Multidimensional prediction of treatment response to antidepressants with cognitive control and functional MRI. Brain. 2017;140:472–86.

Marquand AF, Mourão-Miranda J, Brammer MJ, Cleare AJ, Fu CH. Neuroanatomy of verbal working memory as a diagnostic biomarker for depression. Neuroreport. 2008;19:1507–11.

Costafreda SG, Khanna A, Mourao-Miranda J, Fu CH. Neural correlates of sad faces predict clinical remission to cognitive behavioural therapy in depression. Neuroreport. 2009;20:637–41.

Cao B, Luo Q, Fu Y, Du L, Qiu T, Yang X, et al. Predicting individual responses to the electroconvulsive therapy with hippocampal subfield volumes in major depression disorder. Sci Rep. 2018;8:5434.

Costafreda SG, Chu C, Ashburner J, Fu CH. Prognostic and diagnostic potential of the structural neuroanatomy of depression. PLoS One. 2009;4:e6353.

Wade BSC, Sui J, Njau S, Leaver AM, Vasvada M, Gutman BA, et al. DATA-DRIVEN CLUSTER SELECTION FOR SUBCORTICAL SHAPE AND CORTICAL THICKNESS PREDICTS RECOVERY FROM DEPRESSIVE SYMPTOMS. Proc IEEE Int Symp Biomed Imaging 2017;2017:502–06.

PubMed   PubMed Central   Google Scholar  

Zhang F, Wang C, Lan X, Li W, Ye Y, Liu H, et al. Ketamine-induced hippocampal functional connectivity alterations associated with clinical remission in major depression. J Affect Disord. 2023;325:534–41.

Zhang F, Wang C, Lan X, Li W, Fu L, Ye Y, et al. The functional connectivity of the middle frontal cortex predicts ketamine’s outcome in major depressive disorder. Front Neurosci. 2022;16:956056.

Redlich R, Opel N, Grotegerd D, Dohm K, Zaremba D, Bürger C, et al. Prediction of Individual Response to Electroconvulsive Therapy via Machine Learning on Structural Magnetic Resonance Imaging Data. JAMA Psychiatry. 2016;73:557–64.

Widge AS, Bilge MT, Montana R, Chang W, Rodriguez CI, Deckersbach T, et al. Electroencephalographic Biomarkers for Treatment Response Prediction in Major Depressive Illness: A Meta-Analysis. Am J Psychiatry. 2019;176:44–56.

Wagner S, Tadic A, Roll SC, Engel A, Dreimueller N, Engelmann J, et al. A combined marker of early non-improvement and the occurrence of melancholic features improve the treatment prediction in patients with Major Depressive Disorders. J Affect Disord. 2017;221:184–91.

Lin E, Kuo P-H, Liu Y-L, Yu YWY, Yang AC, Tsai S-J. A Deep Learning Approach for Predicting Antidepressant Response in Major Depression Using Clinical and Genetic Biomarkers. Front Psychiatry. 2018;9:290.

Joyce JB, Grant CW, Liu D, MahmoudianDehkordi S, Kaddurah-Daouk R, Skime M, et al. Multi-omics driven predictions of response to acute phase combination antidepressant therapy: a machine learning approach with cross-trial replication. Transl Psychiatry. 2021;11:513.

Whitfield-Gabrieli S, Ford JM. Default mode network activity and connectivity in psychopathology. Annu Rev Clin Psychol. 2012;8:49–76.

Rolls ET. The cingulate cortex and limbic systems for emotion, action, and memory. Brain Struct Funct. 2019;224:3001–18.

Höflich A, Michenthaler P, Kasper S, Lanzenberger R. Circuit Mechanisms of Reward, Anhedonia, and Depression. Int J Neuropsychopharmacol. 2019;22:105–18.

Wu QZ, Li DM, Kuang WH, Zhang TJ, Lui S, Huang XQ, et al. Abnormal regional spontaneous neural activity in treatment-refractory depression revealed by resting-state fMRI. Hum Brain Mapp. 2011;32:1290–9.

Guo W, Liu F, Xue Z, Gao K, Liu Z, Xiao C, et al. Abnormal resting-state cerebellar-cerebral functional connectivity in treatment-resistant depression and treatment sensitive depression. Prog Neuropsychopharmacol Biol Psychiatry. 2013;44:51–7.

Ye Y, Wang C, Lan X, Li W, Fu L, Zhang F, et al. Baseline patterns of resting functional connectivity within posterior default-mode intranetwork associated with remission to antidepressants in major depressive disorder. Neuroimage Clin. 2022;36:103230.

Xiao H, Yuan M, Li H, Li S, Du Y, Wang M, et al. Functional connectivity of the hippocampus in predicting early antidepressant efficacy in patients with major depressive disorder. J Affect Disord. 2021;291:315–21.

Yamamura T, Okamoto Y, Okada G, Takaishi Y, Takamura M, Mantani A, et al. Association of thalamic hyperactivity with treatment-resistant depression and poor response in early treatment for major depression: a resting-state fMRI study using fractional amplitude of low-frequency fluctuations. Transl Psychiatry. 2016;6:e754.

Hou Z, Gong L, Zhi M, Yin Y, Zhang Y, Xie C, et al. Distinctive pretreatment features of bilateral nucleus accumbens networks predict early response to antidepressants in major depressive disorder. Brain Imaging Behav. 2018;12:1042–52.

Lemke H, Romankiewicz L, Förster K, Meinert S, Waltemate L, Fingas SM, et al. Association of disease course and brain structural alterations in major depressive disorder. Depress Anxiety. 2022;39:441–51.

Sheng W, Cui Q, Jiang K, Chen Y, Tang Q, Wang C, et al. Individual variation in brain network topology is linked to course of illness in major depressive disorder. Cereb Cortex. 2022;32:5301–10.

Fried EI. Moving forward: how depression heterogeneity hinders progress in treatment and research. Expert Rev Neurother. 2017;17:423–25.

Runge VM, Heverhagen JT. The Clinical Utility of Magnetic Resonance Imaging According to Field Strength, Specifically Addressing the Breadth of Current State-of-the-Art Systems, Which Include 0.55 T, 1.5 T, 3 T, and 7 T. Invest Radio. 2022;57:1–12.

Li Q, Zhao Y, Hu Y, Liu Y, Wang Y, Zhang Q, et al. Linked patterns of symptoms and cognitive covariation with functional brain controllability in major depressive disorder. EBioMedicine. 2024;106:105255.

Ai H, Opmeer EM, Marsman JC, Veltman DJ, van der Wee NJA, Aleman A, et al. Longitudinal brain changes in MDD during emotional encoding: effects of presence and persistence of symptomatology. Psychol Med. 2020;50:1316–26.

Li Q, Yao L, You W, Liu J, Deng S, Li B, et al. Controllability of Functional Brain Networks and Its Clinical Significance in First-Episode Schizophrenia. Schizophr Bull. 2023;49:659–68.

Ai Y, Li F, Hou Y, Li X, Li W, Qin K, et al. Differential cortical gray matter changes in early- and late-onset patients with amyotrophic lateral sclerosis. Cereb Cortex. 2024;34:bhad426.

Feng Y, Murphy MC, Hojo E, Li F, Roberts N. Magnetic Resonance Elastography in the Study of Neurodegenerative Diseases. J Magn Reson Imaging. 2024;59:82–96.

Wang Y, Li Q, Yao L, He N, Tang Y, Chen L, et al. Shared and differing functional connectivity abnormalities of the default mode network in mild cognitive impairment and Alzheimer’s disease. Cereb Cortex. 2024;34:bhae094.

You W, Li Q, Chen L, He N, Li Y, Long F, et al. Common and distinct cortical thickness alterations in youth with autism spectrum disorder and attention-deficit/hyperactivity disorder. BMC Med. 2024;22:92.

Hahn A, Wadsak W, Windischberger C, Baldinger P, Höflich AS, Losak J, et al. Differential modulation of the default mode network via serotonin-1A receptors. Proc Natl Acad Sci USA. 2012;109:2619–24.

Chen F, Madsen TM, Wegener G, Nyengaard JR. Repeated electroconvulsive seizures increase the total number of synapses in adult male rat hippocampus. Eur Neuropsychopharmacol. 2009;19:329–38.

Rose D, Fleischmann P, Wykes T, Leese M, Bindman J. Patients’ perspectives on electroconvulsive therapy: systematic review. Bmj. 2003;326:1363.

Wade BS, Joshi SH, Njau S, Leaver AM, Vasavada M, Woods RP, et al. Effect of Electroconvulsive Therapy on Striatal Morphometry in Major Depressive Disorder. Neuropsychopharmacology. 2016;41:2481–91.

Fisher PM, Ozenne B, Ganz M, Frokjaer VG, Dam VN, Penninx BW, et al. Emotional faces processing in major depressive disorder and prediction of antidepressant treatment response: A NeuroPharm study. J Psychopharmacol. 2022;36:626–36.

Brakemeier E-L, Luborzewski A, Danker-Hopfe H, Kathmann N, Bajbouj M. Positive predictors for antidepressive response to prefrontal repetitive transcranial magnetic stimulation (rTMS). J Psychiatr Res. 2007;41:395–403.

Brakemeier E-L, Wilbertz G, Rodax S, Danker-Hopfe H, Zinka B, Zwanzger P, et al. Patterns of response to repetitive transcranial magnetic stimulation (rTMS) in major depression: Replication study in drug-free patients. J Affect Disord. 2008;108:59–70.

Lin H-S, Lin C-H. Early improvement in HAMD-17 and HAMD-6 scores predicts ultimate response and remission for depressed patients treated with fluoxetine or ECT. J Affect Disord. 2019;245:91–97.

Mohamed AK, Croarkin PE, Jha MK, Voort JLV. Early reduction in irritability is associated with improved outcomes among youth with depression: Findings from the AMOD study. J Affect Disord. 2023;324:77–81.

Leaver AM, Wade B, Vasavada M, Hellemann G, Joshi SH, Espinoza R, et al. Fronto-Temporal Connectivity Predicts ECT Outcome in Major Depression. Front Psychiatry. 2018;9:92.

Drysdale AT, Grosenick L, Downar J, Dunlop K, Mansouri F, Meng Y, et al. Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nat Med. 2017;23:28–38.

Kautzky A, Baldinger-Melich P, Kranz GS, Vanicek T, Souery D, Montgomery S, et al. A New Prediction Model for Evaluating Treatment-Resistant Depression. J Clin Psychiatry. 2017 Feb;78:215–+.

Perlis RH. A Clinical Risk Stratification Tool for Predicting Treatment Resistance in Major Depressive Disorder. Biol Psychiatry. 2013;74:7–14.

Su L, Zhang Y, Jia Y, Sun J, Mellor D, Yuan T-F, et al. Predictors of Electroconvulsive Therapy Outcome in Major Depressive Disorder. Int J Neuropsychopharmacol. 2023;26:53–60.

Rezaei M, Bagheri MMS, Ahmadi M. Clinical and demographic predictors of response to anodal tDCS treatment in major depression disorder (MDD). J Psychiatr Res. 2021;138:68–74.

Donse L, Padberg F, Sack AT, Rush AJ, Arns M. Simultaneous rTMS and psychotherapy in major depressive disorder: Clinical outcomes and predictors from a large naturalistic study. Brain Stimul. 2018;11:337–45.

Xue SW, Kuai C, Xiao Y, Zhao L, Lan Z. Abnormal Dynamic Functional Connectivity of the Left RostrHippocampus in Predicting Antidepressant Efficacy in Major Depressive Disorder. Psychiatry Investig. 2022;19:562–69.

Wu H, Liu R, Zhou J, Feng L, Wang Y, Chen X, et al. Prediction of remission among patients with a major depressive disorder based on the resting-state functional connectivity of emotion regulation networks. Transl Psychiatry. 2022;12:391.

Pei C, Sun Y, Zhu J, Wang X, Zhang Y, Zhang S, et al. Ensemble Learning for Early-Response Prediction of Antidepressant Treatment in Major Depressive Disorder. J Magn Reson Imaging. 2020;52:161–71.

Tian S, Sun Y, Shao J, Zhang S, Mo Z, Liu X, et al. Predicting escitalopram monotherapy response in depression: The role of anterior cingulate cortex. Hum Brain Mapp. 2020;41:1249–60.

Zhu J, Cai H, Yuan Y, Yue Y, Jiang D, Chen C, et al. Variance of the global signal as a pretreatment predictor of antidepressant treatment response in drug-naïve major depressive disorder. Brain Imaging Behav. 2018;12:1768–74.

Cash RFH, Cocchi L, Anderson R, Rogachov A, Kucyi A, Barnett AJ, et al. A multivariate neuroimaging biomarker of individual outcome to transcranial magnetic stimulation in depression. Hum Brain Mapp. 2019;40:4618–29.

Ge R, Downar J, Blumberger DM, Daskalakis ZJ, Vila-Rodriguez F. Functional connectivity of the anterior cingulate cortex predicts treatment outcome for rTMS in treatment-resistant depression at 3-month follow-up. Brain Stimul. 2020;13:206–14.

Sun H, Jiang R, Qi S, Narr KL, Wade BS, Upston J, et al. Preliminary prediction of individual response to electroconvulsive therapy using whole-brain functional magnetic resonance imaging data. Neuroimage Clin. 2020;26:102080.

Meyer BM, Rabl U, Huemer J, Bartova L, Kalcher K, Provenzano J, et al. Prefrontal networks dynamically related to recovery from major depressive disorder: a longitudinal pharmacological fMRI study. Transl Psychiatry. 2019;9:64.

Goldstein-Piekarski AN, Korgaonkar MS, Green E, Suppes T, Schatzberg AF, Hastie T, et al. Human amygdala engagement moderated by early life stress exposure is a biobehavioral target for predicting recovery on antidepressants. Proc Natl Acad Sci USA. 2016;113:11955–60.

Williams LM, Korgaonkar MS, Song YC, Paton R, Eagles S, Goldstein-Piekarski A, et al. Amygdala Reactivity to Emotional Faces in the Prediction of General and Medication-Specific Responses to Antidepressant Treatment in the Randomized iSPOT-D Trial. Neuropsychopharmacology. 2015;40:2398–408.

Korgaonkar MS, Williams LM, Song YJ, Usherwood T, Grieve SM. Diffusion tensor imaging predictors of treatment outcomes in major depressive disorder. Br J Psychiatry. 2014;205:321–8.

Queirazza F, Fouragnan E, Steele JD, Cavanagh J, Philiastides MG. Neural correlates of weighted reward prediction error during reinforcement learning classify response to cognitive behavioral therapy in depression. Sci Adv. 2019;5:eaav4962.

Gong Q, Wu Q, Scarpazza C, Lui S, Jia Z, Marquand A, et al. Prognostic prediction of therapeutic response in depression using high-field MR imaging. Neuroimage. 2011;55:1497–503.

Nouretdinov I, Costafreda SG, Gammerman A, Chervonenkis A, Vovk V, Vapnik V, et al. Machine learning classification with confidence: application of transductive conformal predictors to MRI-based diagnostic and prognostic markers in depression. Neuroimage. 2011;56:809–13.

Xu J, Li W, Bai T, Li J, Zhang J, Hu Q, et al. Volume of hippocampus-amygdala transition area predicts outcomes of electroconvulsive therapy in major depressive disorder: high accuracy validated in two independent cohorts. Psychol Med. 2022;53:4464–73.

Download references

This study was supported by the National Key R&D Program (2022YFC2009900) and National Natural Science Foundation (Grant Nos. 82001795 to Youjin Zhao, and 82027808 to Qiyong Gong) of China, and Sichuan Science and Technology Program (2024NSFSC0653 to Fei Li).

Author information

These authors contributed equally: Fenghua Long, Yufei Chen, Qian Zhang.

Authors and Affiliations

Department of Radiology and Huaxi MR Research Center (HMRRC), Functional and Molecular Imaging Key Laboratory of Sichuan Province, West China Hospital, Sichuan University, Chengdu, 610041, Sichuan Province, PR China

Fenghua Long, Yufei Chen, Qian Zhang, Qian Li, Yaxuan Wang, Yitian Wang, Haoran Li, Youjin Zhao, John A. Sweeney, Qiyong Gong & Fei Li

Research Unit of Psychoradiology, Chinese Academy of Medical Sciences, Chengdu, 610041, Sichuan, PR China

Fenghua Long, Yufei Chen, Qian Zhang, Qian Li, Yaxuan Wang, Yitian Wang, Haoran Li, Youjin Zhao, Qiyong Gong & Fei Li

Department of Psychiatry and Behavioral Neuroscience, University of Cincinnati, Cincinnati, OH, 45219, USA

Robert K. McNamara, Melissa P. DelBello & John A. Sweeney

You can also search for this author in PubMed   Google Scholar

Contributions

Fei Li conceptualized the study. Fenghua Long, Yufei Chen, Qian Zhang, John A. Sweeney, and Fei Li designed and drafted the manuscript. Fenghua Long, Yufei Chen, Qian Zhang, Youjin Zhao, Qian Li, Yitian Wang, Haoran Li, and Yaxuan Wang contributed to the literature search, data collection and analysis. Fenghua Long, Yufei Chen, Qian Zhang, Robert K. McNamara, Melissa P. DelBello, John A. Sweeney, Qiyong Gong, and Fei Li critically revised the manuscript, and contributed to the data analysis strategy and interpretation. All authors approved the final version of the manuscript.

Corresponding author

Correspondence to Fei Li .

Ethics declarations

Competing interests.

Melissa P. DelBello receives research support from national institutes of health (NIH), PCORI, Acadia, Allergan, Janssen, Johnson and Johnson, Lundbeck, Otsuka, Pfizer, and Sunovion. She is also a consultant, on the advisory board, or has received honoraria for speaking for Alkermes, Allergan, Assurex, CMEology, Janssen, Johnson and Johnson, Lundbeck, Myriad, Neuronetics, Otsuka, Pfizer, Sunovion, and Supernus. Robert K. McNamara has received research support from Martek Biosciences Inc, Royal DSM Nutritional Products, LLC, Inflammation Research Foundation, Ortho-McNeil Janssen, AstraZeneca, Eli Lilly, NARSAD, and NIH, and previously served on the scientific advisory board of the Inflammation Research Foundation. The remaining authors declare no potential conflicts of interest with regard to this manuscript.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary materials, rights and permissions.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Cite this article.

Long, F., Chen, Y., Zhang, Q. et al. Predicting treatment outcomes in major depressive disorder using brain magnetic resonance imaging: a meta-analysis. Mol Psychiatry (2024). https://doi.org/10.1038/s41380-024-02710-6

Download citation

Received : 18 January 2024

Revised : 18 August 2024

Accepted : 19 August 2024

Published : 26 August 2024

DOI : https://doi.org/10.1038/s41380-024-02710-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

ABC Imaging

A publication of the Cardiovascular Imaging Department of the Brazilian Society of Cardiology

Uma publicação do Departamento de Imagem Cardiovascular da Sociedade Brasileira de Cardiologia

Search Articles

Arq bras cardiol: imagem cardiovasc. 2024; 37(3): e20240044, my approach to cardiovascular computed tomography and magnetic resonance imaging in the evaluation of cardiac pseudotumors: a brief review.

DOI: 10.36660/abcimg.20240044i

Cardiac pseudotumoral lesions are non-neoplastic conditions that are often overlooked in the differential diagnosis of cardiac masses. They present a variable clinical picture, ranging from asymptomatic to causing complications such as ventricular filling restriction and outflow tract obstruction. Echocardiography is the first-line imaging method but has limitations, such as dependence on the acoustic window and operator variability. However, a multimodality approach, including CT and MRI, is essential for seeking an accurate diagnosis. CT, with its excellent spatial resolution, allows for anatomical detailing, assessment of intralesional calcifications and fat, and contributes to therapeutic planning. MRI is preferred for tissue characterization and differentiation between benign and malignant lesions. Normal anatomical structures, such as the Eustachian valve and Chiari network, can be confused with thrombi or tumors, requiring correct identification. Thrombi are common in patients with atrial fibrillation or mitral valve disease, with MRI being important for differentiating them from neoplasms. Other pseudotumoral conditions include vegetation, gossypibomas, caseous calcification of the mitral valve annulus, and lipomatous hypertrophy of the interatrial septum. The integration of advanced cardiovascular imaging modalities is fundamental for the diagnosis and management of these lesions, optimizing patient care.

Keywords: Emission-Computed Tomography ; Heart neoplasms ; Magnetic Resonance Spectroscopy

How to cite this article

Torres R V A , Fuzissima B M , Fonseca E K U N , Farias L P G , Ishikawa W Y , Araújo Filho J A B . My Approach to Cardiovascular Computed Tomography and Magnetic Resonance Imaging in the Evaluation of Cardiac Pseudotumors: A Brief Review. Arq Bras Cardiol: Imagem cardiovasc. 2024;37(3):e20240044.

Torres, Roberto Vitor Almeida ; Fuzissima, Bruno Maeda ; Fonseca, Eduardo Kaiser Ururahy Nunes ; Farias, Lucas de Pádua Gomes de ; Ishikawa, Walther Yoshiharu ; Araújo Filho, José de Arimatéia Batista . My Approach to Cardiovascular Computed Tomography and Magnetic Resonance Imaging in the Evaluation of Cardiac Pseudotumors: A Brief Review. Arq Bras Cardiol: Imagem cardiovasc. , v. 37, n. 3, e20240044, Aug. 2024.

Torres, R. V. A. , Fuzissima, B. M. , Fonseca, E. K. U. N. , Farias, L. P. G. , Ishikawa, W. Y. , & Araújo Filho, J. A. B. (2024). My Approach to Cardiovascular Computed Tomography and Magnetic Resonance Imaging in the Evaluation of Cardiac Pseudotumors: A Brief Review. Arq Bras Cardiol: Imagem cardiovasc., 37 (3), e20240044.

Torres, Roberto Vitor Almeida and Fuzissima, Bruno Maeda and Fonseca, Eduardo Kaiser Ururahy Nunes and Farias, Lucas de Pádua Gomes de and Ishikawa, Walther Yoshiharu and Araújo Filho, José de Arimatéia Batista . My Approach to Cardiovascular Computed Tomography and Magnetic Resonance Imaging in the Evaluation of Cardiac Pseudotumors: A Brief Review. Arq Bras Cardiol: Imagem cardiovasc. [online]. 2024, vol. 37, n. 3, [cited 2024-08-29], e20240044. Available from: <https://www.abcimaging.org/article/my-approach-to-cardiovascular-computed-tomography-and-magnetic-resonance-imaging-in-the-evaluation-of-cardiac-pseudotumors-a-brief-review/>. ISSN .

2675-312X-abcimg-37-03-e20240044-gf11

Dataset Informations

Comments cancel reply.

Report of a case of renal collecting duct carcinoma with literature review

• Open access
• Published: 29 August 2024
• Volume 1 , article number  33 , ( 2024 )

Cite this article

You have full access to this open access article

• Yong Cui 1 &
• Yuan Gao 2  

Collecting duct carcinoma (CDC) is a rare pathologic subtype of renal cell carcinoma that accounts for approximately 0.4–1% of cases and originates in the collecting ducts (Bellini ducts) of the renal medulla. The majority of patients are metastatic at the time of presentation, extremely malignant and rapidly progressive, with most patients dying 1–3 years after initial diagnosis. Currently, surgery is considered the only effective treatment. There is no uniform standard for postoperative adjuvant radiotherapy and chemotherapy.In this article, we report a case of an 85-year-old female patient with CDC who underwent radical nephrectomy. This is an extremely rare case. We describe the case and perform a literature review to report current research advances regarding the treatment and prognosis of patients with CDC. The aim of this study is to contribute to improving the diagnosis and treatment of CDC.

Avoid common mistakes on your manuscript.

1 Introduction

CDC is a rare pathologic subtype of renal cell carcinoma that accounts for approximately 0.4–1% of cases and originates from the collecting ducts of the renal medulla (Bellini ducts), Mancilla-Jimenz reported the first case of CDC [ 1 ].Due to the rapid progression of the disease and extensive metastasis to peripheral lymph nodes, the majority of patients have a very poor prognosis, with a median survival of less than 2 years [ 2 ]. Early diagnosis is an important factor in prolonging survival.Fleming and Lewi developed diagnostic criteria and CDC was isolated as a separate subtype of renal cancer.Although this disease has immunohistochemical features, it is difficult to characterize histologically due to its rarity and needs to be differentiated from other renal cell carcinomas and uroepithelial carcinomas [ 3 ].

We retrospectively analyzed the medical records of a patient with CDC admitted to Weifang People’s Hospital, and combined with the literature review, we analyzed the diagnosis and treatment of CDC in order to improve people's understanding of CDC, so as to reduce misdiagnosis and mistreatment.

2 Case report

An 85-year-old woman was admitted to the hospital with a complaint of hematuria for 3 months. The patient reported no distinct triggers or causes for the hematuria. She had chest pain and chest tightness. Prior to admission, she did not have lumbar pain, loss of appetite, fatigue, lethargy, or other systemic symptoms. A lower abdominal and pelvic computed tomography (CT) scan performed after admission revealed the following:Left upper renal calyx and renal parenchyma occupations; Small cyst of the right kidney; pelvic effusion (Fig.  1 ). Considering her age and the necessity of surgery, a team of urologists evaluated his physical condition and established a treatment plan for surgery. One week after admission, the patient underwent laparoscopic nephroureterectomy, which was successful.

The lower abdominal and pelvic CT scan revealed Left upper renal calyx and renal parenchyma occupations

2.1 Postoperative pathological diagnosis

Postoperative pathological examination revealed resection of the left kidney along with perirenal fat measuring 12 cm × 7 cm × 4.5 cm. The renal fat capsule was easily peeled. A mass measuring 2.5 cm × 1.5 cm × 1 cm was observed in the kidney. The section appeared grayish white and grayish red, was brittle, and had a close relationship with the renal peritoneum, 0.6 cm from it.

2.2 Pathological results

The left kidney tumor was glandular tubular and papillary. The size of the renal tumor was 2.5 cm × 1.5 cm × 1 cm. The tumor involved the renal pelvis and renal sinus fat, and did not involve the renal perineurium and perirenal fat. The nerves were invaded and cancerous emboli were seen in the vasculature. Ureteral and vascular breaks were clear (Fig.  2 ).

Pathological findings of renal collecting duct carcinoma (hematoxylin and eosin, 100×)

Immunohistochemical results: CK wide ( +), Vimentin ( +), E-Cadherin (partially +), RCC (scattered +), PAX-8 ( +),CAIX (focal +), P504S ( +), INI1 (weak +), Ki-67 (index 20%).

Pathological staging based on the American Joint Committee on Cancer (AJCC) 8th edition classified the tumor as pT3aNxMx.

2.3 Postoperative treatment

Postoperatively, pezopanib 0.8 g was administered orally once daily.Intravenous 160 mg of teraplizumab every 3 weeks for 1 month postoperatively. CT scans of the lower abdomen at 1 month postoperatively and 3 months postoperatively reported the following findings:

Postoperative left kidney; Striated dense shadow in the left adrenal region.This treatment has been maintained for 4 months. No adverse immune-related events occurred, and no signs of local recurrence or systemic metastasis were found. Therefore, the therapeutic effect has been highly favorable.In April after surgery, the patient was admitted to the hospital for diarrhea, and the patient died 3 days after admission.

3 Discussion

3.1 clinical manifestations.

CDC can occur at any age and is more common in younger people. Males are more likely to be affected than females by a ratio of approximately 2:1 [ 4 ].

Common clinical symptoms of CDC include painless hematuria, low back and abdominal pain, low back and abdominal masses, fatigue, fever, and weight loss or when the tumor develops metastasis, Usually most of the patients develop metastasis including bone and lymph node metastasis prior to treatment [ 5 ]. In this case, we presented to the hospital with painless hematuria as the main symptom.

3.2 Imaging features

The imaging features of CDC differ from those of other renal cell carcinomas. In renal tumors located in the zone connecting the renal cortex and medulla, the early presentation of poorly defined borders, mild enhancement, and metastasis require consideration of the diagnosis of CDC [ 6 ].

The organ of origin of CDC is water-rich renal tissue, and the interstitium of the tumor is characterized by increased fibrous tissue proliferation and collagenization, and the tumor parenchyma is denser than the surrounding normal tissue, which is a characteristic feature of CDC tumors on nonenhanced CT. These features differ from renal cell carcinoma originating in the renal cortex, where dense interstitial tissue or increased collagen secretion, proliferation of inflammatory fibroblast tissue and abundant fibrous tissue components are characteristic of CDC tumors. These features are important for the pathologic diagnosis of CDC, which shows low signal on magnetic resonance T2WI [ 7 ]. On dynamic enhancement scans, most CDC tumors show relatively low density in the renal cortex and medulla. The parenchyma of the mass is heterogeneous, with mild to moderate enhancement in the cortical or medullary stage, below the surrounding renal parenchyma. The medullary stage shows inhomogeneous and mildly delayed enhancement [ 6 ]. MRI shows isointensity or hyperintensity on T1WI and hypointensity on T2WI [ 8 ].

3.3 Pathologic features

Pathological examination is the gold standard for the diagnosis of CDC. Tumors tend to be yellowish-brown to white in color, may be associated with necrosis and hemorrhage, and show infiltrative growth, often invading the cortex, and may extend beyond the renal parenchyma, including perirenal fat, adrenal glands, and perirenal fascia [ 9 ]. Microscopically, CDCs are usually seen as tubular papillary structures, with tumor cells forming cobblestones along the glandular ducts. Lowly differentiated tumor cells show a nested, cord-like, sarcomatoid, or adenoidal cystic morphology with or without interstitial connective tissue reaction [ 10 ].

Genetic and biochemical approaches are becoming increasingly important for the identification and diagnosis of renal cancer. Many biomolecules, including epithelial mesenchymal transition (EMT) markers, such as N-calmodulin and poikilodulin, and human leukocyte molecules, such as HLA-G and HLA-E, have been reported to be biomarkers for renal cancer. Immunohistochemical examination of these biomarkers is important for determining the origin and diagnosis of CDC. positive expression of CK (AE1/AE3), CK7, CK19, EMA, wave proteins, CK34BE12, PNA, and european agglutinin (UEA), and negative expression of CD10 and CK20. the combination of CK34BE12 and PNA is able to detect 90% of the CDC. Pathologic and immunohistochemical findings are important for the diagnosis of CDC and differentiation from other types of renal cancer [ 11 , 12 , 13 ].

3.4 Differential diagnosis

Renal collecting duct carcinoma is mainly differentiated from the following malignant tumors.

Renal medullary carcinoma: renal medullary carcinoma originates from the collecting ducts near the cortical area, under the microscope, the tumor of renal medullary carcinoma is low differentiated and patchy distribution, and the tumor cells are arranged in an adenoidal cystic structure, and more neutrophil infiltration can be seen in the tumor body, and sickle-shaped erythrocytes can be seen at the same time [ 14 ].

Papillary renal cell carcinoma: According to the histopathological changes, it is divided into two subtypes: type I and type II. Tumor cells consist of papillary or tubular structures with a slender vascular axis, and foamy macrophages and cholesterol crystals can be seen in the core of the papillae; tumor cells are small, with sparse cytoplasm (type I) or tumor cells with abundant eosinophilic cytoplasm and high nuclear grading (type II), and areas of necrosis, sarcomatous differentiation and rhabdomyosarcomatous differentiation can be seen. Papillary renal cell carcinoma was positive for CK7, and the positive rate of P504S was high.

3.5 Treatment and prognosis

In the clinical setting, the initial diagnosis and treatment decisions for renal cancer are often made without histopathologic information, and the diagnosis can only be confirmed based primarily on imaging and after initial surgery. According to the literature, most CDC cases are high-grade and advanced, but there is no consensus on treatment options. Surgery remains the most effective treatment for patients with kidney cancer, even in advanced stages [ 15 ]. Surgical procedures include radical nephrectomy and partial nephrectomy, but given the highly aggressive nature of CDC, radical nephrectomy is recommended. Méjean et al. [ 16 ] reported that the median survival of six patients with CDC who underwent nephrectomy was only seven months, and three patients died perioperatively or postoperatively. CDC is highly biologically invasive, and patients are usually in poor general condition at the time of diagnosis. Surgical complications and postoperative recovery may prevent patients from receiving systemic therapy, so it is recommended that the diagnosis of CDC can be confirmed by biopsy, which can provide the best diagnostic and treatment plan for subsequent treatment.

In patients with CDC who develop metastases, the treatment regimen is usually based on uroepithelial tumors, using chemotherapy regimens, usually platinum or carboplatin in combination with gemcitabine. The prognosis for these regimens is poor, with survival statistically less than 12 months [ 17 ]. Orsola et al. [ 18 ] reported two cases of CDC who underwent adjuvant chemotherapy (doxorubicin + gemcitabine) after radical nephrectomy; however, the mean postoperative survival was only 5.6 months. A prospective phase II trial showed that, platinum-based combined with gemcitabine chemotherapy was effective in 23 patients with previously untreated mCDC undergoing 6 cycles of treatment, which showed objective remission rate (ORR), progression-free survival (PFS), and overall survival (OS) of 26 months, 7.1 months, and 10.5 months, respectively [ 19 ].

Tyrosine kinase inhibitors (TKIs) and mTOR inhibitors targeting the vascular endothelial growth factor (VEGF) pathway are important for the treatment of metastatic CCRCC. There is less evidence on the effectiveness of such agents in the treatment of NCCRCC, especially for mCDC.A small retrospective analysis of 13 patients with mCDC evaluated the activity of different TKIs: sunitinib, sorafenib, pazopanib, and the mTOR inhibitor tesirolimus. Only 2 patients were able to receive second-line treatment with sunitinib after disease progression, no patient survived more than 5 years, 4 patients experienced early disease progression, and 2 patients had long-term control of their disease with OS of 49 and 19 months, respectively [ 20 ].

Tamada et al. [ 21 ] reported that immunotherapy-based combination therapy has become the standard first-line treatment for metastatic RCC.CDC in combination with immunotherapy, the effect of which is unclear. A 62-year-old man was treated with pembrolizumab and axitinib for CDC with multiple bone metastases. after 7 months, the primary and metastatic lesions shrank and were evaluated as partial response.A 71-year-old man received pembrolizumab and axitinib for the treatment of CDC with lymph nodes and lung metastases. after 9 months, the primary and metastatic lesions had shrunk and were assessed as partial responses. Tumor cell expression of PD-L was negative in both patients, and CD4 + and CD8 + cells were observed in the tumors. The combination of pembrolizumab and axitinib is effective in the immunotherapy of metastatic CDC [ 22 ].

In summary, despite clinical advances in the treatment of metastatic RCC, the prognosis for patients with mCDC remains poor. Currently available treatment options are poor, and results from prospective trials are limited. For patients with localized disease, nephrectomy is the only potentially curative option, whereas for patients with mCDC, a doublet chemotherapy regimen containing platinum salts and gemcitabine is recommended.

4 Conclusion

In conclusion, CDC is a highly malignant and rare renal tumor with unique biological behavior, morphology and functional manifestations. Most patients with CDC have distant metastases at the time of initial diagnosis. The prognosis of CDC is generally poor. There is no standardized follow-up treatment plan for this tumor, and radical surgery is the only curative treatment for patients with limited CDC. However, it is prone to recurrence and metastasis after surgery.

Data availability

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

Tang C, Zhou Y, Ge S, et al. Incidence, clinical characteristics, and survival of collecting duct carcinoma of the kidney: a population-based study. Front Oncol. 2021;11: 727222.

Article   PubMed   PubMed Central   Google Scholar  

Bahadoram S, Davoodi M, Hassanzadeh S, et al. Renal cell carcinoma: an overview of the epidemiology, diagnosis, and treatment. G Ital Nefrol. 2022;39(3):9.

Google Scholar  

Panunzio A, Tappero S, Hohenhorst L, et al. Collecting duct carcinoma: epidemiology, clinical characteristics and survival. Urol Oncol. 2023;41(2):110–7.

Article   Google Scholar  

Tokuda N, Naito S, Matsuzaki O, et al. Collecting duct (Bellini duct) renal cell carcinoma: a nationwide survey in Japan. J Urol. 2006;176(1):40–3.

Article   PubMed   Google Scholar  

Wang X, Hao J, Zhou R, et al. Collecting duct carcinoma of the kidney: a clinicopathological study of five cases. Diagn Pathol. 2013;8:96.

Lyu Z, Liu L, Li H, et al. Imaging analysis of 13 rare cases of renal collecting (Bellini) duct carcinoma in northern China: a case series and literature review. BMC Med Imaging. 2021;21(1):42.

Zhu Q, Wu J, Wang Z, et al. The MSCT and MRI findings of collecting duct carcinoma. Clin Radiol. 2013;68(10):1002–7.

Article   CAS   PubMed   Google Scholar  

Pickhardt PJ, Siegel CL, McLarney JK. Collecting duct carcinoma of the kidney: are imaging findings suggestive of the diagnosis? AJR Am J Roentgenol. 2001;176(3):627–33.

Seo AN, Yoon G, Ro JY. Clinicopathologic and molecular pathology of collecting duct carcinoma and related renal cell carcinomas. Adv Anat Pathol. 2017;24(2):65–77.

Chao D, Zisman A, Pantuck AJ, et al. Collecting duct renal cell carcinoma: clinical study of a rare tumor. J Urol. 2002;167(1):71–4.

Li M, Vuolo MA, Weidenheim KM, et al. Collecting-duct carcinoma of the kidney with prominent signet ring cell features. Mod Pathol. 2001;14(6):623–8.

Kren L, Valkovsky I, Dolezel J, et al. HLA-G and HLA-E specific mRNAs connote opposite prognostic significance in renal cell carcinoma. Diagn Pathol. 2012;7:58.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Behnes CL, Hemmerlein B, Strauss A, et al. N-cadherin is differentially expressed in histological subtypes of papillary renal cell carcinoma. Diagn Pathol. 2012;7:95.

Elliott A, Bruner E. Renal medullary carcinoma. Arch Pathol Lab Med. 2019;143(12):1556–61.

Guillaume Z, Allory Y, Auclin E, et al. Collecting duct carcinoma and renal medullary carcinoma in the age of new therapies. Bull Cancer. 2023;110(4):450–62.

Méjean A, Rouprêt M, Larousserie F, et al. Is there a place for radical nephrectomy in the presence of metastatic collecting duct (Bellini) carcinoma? J Urol. 2003;169(4):1287–90.

Panunzio A, Sorce G, Tappero S, et al. Mortality according to treatment in metastatic collecting duct renal cell carcinoma. Clin Genitourin Cancer. 2023;21(2):295–300.

Orsola A, Trias I, Raventós CX, et al. Renal collecting (Bellini) duct carcinoma displays similar characteristics to upper tract urothelial cell carcinoma. Urology. 2005;65(1):49–54.

Oudard S, Banu E, Vieillefond A, et al. Prospective multicenter phase II study of gemcitabine plus platinum salt for metastatic collecting duct carcinoma: results of a GETUG (Groupe d’Etudes des Tumeurs Uro-Génitales) study. J Urol. 2007;177(5):1698–702.

Procopio G, Testa I, Iacovelli R, et al. Treatment of collecting duct carcinoma: current status and future perspectives. Anticancer Res. 2014;34(2):1027–30.

PubMed   Google Scholar  

Tamada S, Ikarashi D, Tsuyukubo T, et al. Efficacy of combination therapy with pembrolizumab and axitinib for metastatic renal collecting duct cell carcinoma: a report on two cases. IJU Case Rep. 2022;5(6):438–41.

Plimack ER, Powles T, Stus V, et al. Pembrolizumab plus axitinib versus sunitinib as first-line treatment of advanced renal cell carcinoma: 43-month follow-up of the phase 3 KEYNOTE-426 study. Eur Urol. 2023;84(5):449–54.

Download references

Author information

Authors and affiliations.

Department of Clinical Medicine, Shandong Second Medical University, Weifang, Shandong, China

Department of Urinary Surgery, Weifang People’s Hospital, No.151 Guangwen Street, Weifang, 261000, Shandong, China

You can also search for this author in PubMed   Google Scholar

Contributions

C.Y: Writing—original draft. G.Y and C.Y: Writing—review and editing.

Corresponding author

Correspondence to Yuan Gao .

Ethics declarations

Ethics approval and consent to participate.

This study was approved by the Ethics Committee of Weifang People’s Hospital. We certify that the study was performed in accordance with the 1964 declaration of HELSINKI and later amendments. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Cui, Y., Gao, Y. Report of a case of renal collecting duct carcinoma with literature review. Discov Med 1 , 33 (2024). https://doi.org/10.1007/s44337-024-00051-5

Download citation

Received : 19 June 2024

Accepted : 26 August 2024

Published : 29 August 2024

DOI : https://doi.org/10.1007/s44337-024-00051-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Renal collecting duct carcinoma
• Case report
• Literature review
• Find a journal
• Publish with us
• Track your research

• Case report
• Open access
• Published: 28 August 2024

Metastasis of malignant melanoma to urinary tract: a case report

• Takashi Ueno 1 ,
• Takumi Kiwaki 2 ,
• Hironori Betsunoh 3 ,
• Kaoru Ito 1 ,
• Takaya Murashima 1 ,
• Masato Fujii 1 ,
• Takahiro Nagai 1 ,
• Shoichiro Mukai   ORCID: orcid.org/0000-0002-1678-8619 1 ,
• Atsuro Sawada 1 &
• Toshiyuki Kamoto 1  

Journal of Medical Case Reports volume  18 , Article number:  396 ( 2024 ) Cite this article

Metrics details

Introduction

Metastasis of malignant melanoma to urinary tract is reported to be rare. According to retrospective analysis of a single center study, improvement of overall survival was observed in patients with metastasis to the gastrointestinal tract that had undergone metastasectomy with curative intent. However, there is no significant evidence regarding resection for metastasis to urinary tract.

Case presentation

Case 1: an 86-year-old Japanese man was diagnosed with a small bladder tumor by computed tomography scan during post operative follow-up of malignant melanoma in the choroid of the left eye. Cystoscopy revealed black, nonpapillary tumors, suggesting metastatic malignant melanoma. Because no apparent invasive growth to muscle layer was observed by magnetic resonance imaging, transurethral resection was performed. Pathological appearance was compatible with metastatic malignant melanoma. No recurrence in urinary tract was observed; however, multiple liver metastasis was diagnosed at 3 months after surgery. Case 2: a 57-year-old Japanese man was diagnosed with right hydronephrosis due to ureteral tumor. He had a past history of subungual malignant melanoma to the left thumb 2 years prior to his visit. Right nephroureterectomy was performed, and pathological evaluation revealed metastatic malignant melanoma. He revisited 2 years later due to dysuria, and a large bladder tumor was revealed by ultrasound. Cystoscopy showed black-colored nonpapillary tumor, suggesting malignant melanoma. Total cystectomy was recommended; however, the patient withheld consent. Therefore, we performed transurethral resection. The resulting pathological finding was compatible with metastatic malignant melanoma without invasion to muscle layer. He remained free from local recurrence and metastasis for 22 years after surgery.

We successfully performed metastasectomy for bladder and ureteral metastases without recurrence in the urinary tract. Long recurrence-free survival was observed in case 2. Complete resection for metastasis of malignant melanoma may have the potential to improve survival.

Peer Review reports

Malignant melanoma commonly occurs in the skin and metastasizes to lymph nodes (42–59%), lungs (18–36%), and liver (14–20%) [ 1 ]. Although bladder metastasis of malignant melanoma is reported in 14–22% of autopsy cases [ 2 ], clinically apparent cases reported in the English-language literature accounted for 31 cases, suggesting that metastatic lesions to the urinary tract are difficult to detect while patients are alive because they are less likely to exhibit symptoms [ 3 , 4 , 5 ]. Here, we report two cases of malignant melanoma metastasizing to urinary tract that were successfully treated with surgical intervention.

An 86-year-old Japanese man underwent left-eye enucleation due to malignant melanoma of the choroid 4 years prior to his visit. After surgery, computed tomography (CT) scan was performed every 6 months to screen for metastasis, with the most recent contrast CT scan revealing a small tumor in the bladder with weak enhancement (Fig.  1 A). The patient was then referred to our department, where cystoscopy was performed. The examination showed two black, nonpapillary tumors on the anterior wall of the bladder measuring approximately 5 mm and less than 1 mm, respectively (Fig.  1 B). Cytology of the urine was negative. MRI revealed small tumor showing high intensity on T1WI and low intensity on T2WI at anterior bladder wall without apparent invasive growth into the muscle layer (Fig.  1 C). Since the size of the tumors was small and less likely to invade the muscle layer, we decided to perform transurethral resection of bladder tumor (TUR-Bt). In the surgery, we incised the bladder mucosa from approximately 1 cm away from the tumors. Histological examination revealed proliferation of atypical cells with rounded hyperchromatic nuclei and intracytoplasmic melanin granules with a nodular appearance at the submucosal layer (Fig.  1 D, E). Immunohistochemically, the tumor cells were positive for Melan A and HMB-45 (Fig.  1 F). The appearance was similar to primary melanoma (data not shown) and compatible with metastasis of malignant melanoma. No apparent muscular invasion was observed.

Clinical and pathological appearance of case 1. Sagittal imaging of contrast computed tomography (nephrogenic phase) is shown ( A ). A small tumor with weak enhancement is suggested on the anterior side of the bladder. Cystoscopic appearance shows two nonpapillary black tumors ( B ). The bladder tumor shows high intensity in T1 weighted image (left), low intensity in T2 weighted image (right). No apparent muscle invasion is revealed in T2 weighted image of magnetic resonance imaging examination ( C ). Histological findings are shown ( D – F ). Atypical cells with intracytoplasmic melanin granules proliferate in the submucosal layer ( D , E ). The tumor cells are positive for Melan A (counterstaining with Giemsa, F )

The patient recovered without postoperative events. Three months after TUR-Bt, multiple liver metastases were discovered on CT scan. Since genetic analysis revealed no mutation in BRAF , treatment by immune checkpoint inhibitor was considered. However, the patient chose best supportive care without additional medical treatment. The general condition remained stable without recurrence in urinary tract at 6 months after surgery.

A 57-year-old Japanese man visited a private clinic with the chief complaint of right abdominal discomfort. Ultrasound examination revealed right hydronephrosis. Because CT scan suggested that the hydronephrosis was due to ureteral tumor (Fig.  2 A), the patient was referred to our department. The patient had undergone surgical intervention for subungual malignant melanoma to the left thumb 2 years prior to his visit. Retrograde pyelography showed complete obstruction of upper ureter, and the obstructed portion was matched to the ureteral tumor diagnosed by CT scan. Although urine cytology was negative, CT scan findings were compatible with ureteral carcinoma; therefore, we performed right nephroureterectomy. The resected specimen showed yellow–white and partially black colored pedunculated tumor to the upper ureter, and no satellite tumors were evident in the urinary tract (Fig.  2 B). Pathological findings showed submucosal proliferation of polygonal to spindle-shaped cells with oval to irregular-shaped nuclei and intracytoplasmic melanin granules (Fig.  2 C). The tumor cells were positive for Melan A and HMB-45, prompting a diagnosis of ureteral metastasis of known malignant melanoma. The patient recovered without postoperative event. A month after surgery, full-body radiological examination by CT scan, MRI, and bone scintigraphy was performed. Results revealed no apparent metastasis. Two years later, the patient revisited with a complaint of dysuria. Ultrasound and CT scan showed bladder tumor measuring approximately 4 cm in diameter (Fig.  2 D), and cystoscopy revealed black-colored nonpapillary tumor, suggesting malignant melanoma (Fig.  2 E). Because the tumor was large and invasion could not be ruled out, we recommended total cystectomy; however, the patient withheld consent. Therefore, we performed transurethral resection of bladder tumor (TUR-Bt). Pathological diagnosis of the resected specimens was compatible with malignant melanoma similar to the previously resected ureteral tumor (Fig.  2 F–H). Fortunately, no apparent invasion to muscle layer was observed, and the patient remained free from local recurrence and metastasis for 22 years after TUR-Bt.

Clinical and pathological appearance of case 2. Contrast computed tomography (early phase) reveals solid mass with weak enhancement in the upper ureter ( A , white arrow). Macroscopic appearance (cutting surface) of resected right kidney and the ureter is shown ( B ). Yellow–white and partially black colored pedunculated tumor in the upper ureter (yellow arrow). Contrast computed tomography (excretory phase) shows large bladder tumor ( D ), and cystoscopy reveals black-colored nonpapillary tumor ( E ). Histological appearances of ureteral tumor ( C ) and bladder ( F – H ) show submucosal proliferation of polygonal to spindle-shaped cells with oval nuclei and intracytoplasmic melanin granules. The tumor cells were positive for Melan A

Secondary bladder neoplasms have been reported to represent 2–3% of all malignant bladder tumors; however, the majority of these spread directly from adjacent organs, including colon, prostate, rectum, and cervix [ 6 , 7 ]. In metastatic secondary bladder neoplasm, the most common primary site is stomach (4.3% of all secondary bladder neoplasms), followed by skin (3.9%), lung (2.8%), and breast (2.5%) [ 6 , 7 , 8 ]. The most common histological type of secondary bladder neoplasm has been reported to be adenocarcinoma, with malignant melanoma being less common. However, a higher incidence of bladder metastasis of malignant melanoma was reported in autopsy cases. Sheehan et al . analyzed 5200 autopsy cases and reported 21 cases of metastatic bladder tumors, with 8 of these (38%) being metastatic malignant melanomas [ 9 ]. Bates reported as a possible reason for the frequency of secondary tumors to urinary and male genital tracts being higher in autopsy cases that autopsy cases were more likely to have disseminated disease [ 10 ]. In addition, sampling bias toward unusual lesions and convenience of observation by autopsy for outer layer of hollow organ were discussed with the conclusion that it is simpler to diagnose secondary neoplasia at autopsy than on the basis of biopsy.

Metastatic malignant melanoma of the bladder is typically reported to present as asymptomatic macroscopic hematuria and diagnosed by cystoscopy and histopathological features considered with clinical history of previous melanoma. Diagnostic criteria to determine whether malignant melanoma of the bladder is a primary tumor include (1) absence of any previous skin lesions, (2) absence of cutaneous malignant melanoma, (3) absence of primary visceral malignant melanoma, (4) absence of recurrence pattern showing consistency with the primary tumor diagnosis, and (5) atypical melanocytes at the tumor margin upon microscopic examination [ 9 ]. In our two cases, each patient had an apparent past history of primary melanoma, and the bladder tumor was pathologically compatible with primary tumor. Therefore, we diagnosed metastatic malignant melanoma of the bladder in both cases.

Ureteral metastasis of malignant melanoma is extremely rare. To the best to our knowledge, this is the ninth case report [ 11 , 12 , 13 , 14 ]. In addition, malignant melanoma occurred in the bladder metachronously. Because the tumor was located in the submucosal region (not the superficial region), metastasis was considered rather than intraluminal seeding from ureteral melanoma. Although all tumors were located in the submucosal area, no apparent muscle invasion was observed. Therefore, the tumors were completely resected without local recurrence.

When distant multiple metastases are discovered in patients with malignant melanoma, surgical intervention may be less indicated. However, complete resection may have the potential to improve overall survival (OS) in some cases. Deutsch et al . reported the patients with abdominal visceral metastases undergoing surgical resection had superior overall survival compared with patients treated with medical agents only, including new agents (18 months versus 7 months; P  < 0.001) [ 15 ]. The study included 366 cases with metastasis in the gastrointestinal tract, 697 cases in the liver, 138 cases in the adrenal glands, 38 cases in the pancreas, 109 cases in the spleen, and 305 cases with multiple sites. Patients with metastasis to the gastrointestinal tract undergoing complete curative resection had the greatest benefit from metastasectomy (median OS of 64 months). However, no apparent benefit for OS was observed in patients receiving palliative surgery. The study included a large number of cases and yielded significant results; however, a limitation is that it was a retrospective analysis. Further prospective study is recommended to clarify the significance of metastasectomy.

Although there was no metastasis in another organs at surgery (complete metastasectomy was performed in this period), liver metastasis appeared at 3 months after TUR-Bt in case 1. On the other hand, long-term disease control (22 years) was observed in case 2 by complete resection. According to literature, complete curative metastasectomy may have benefit for patient survival, especially in metastasis to gastrointestinal tract; however, there was no evidence of metastasectomy in patients with metastasis to urinary tract. Therefore, accumulation and analysis of the cases with urinary tract metastasis will be necessary to clarify the benefit of metastasectomy.

On contrast CT scan, metastatic melanoma usually manifested as hyper vascular mass [ 16 ]. MRI showed high intensity on T1WI and low intensity mass on T2WI [ 17 ]. In our cases, weak enhancement was observed by contrast CT scan in both cases. In addition, MRI appearance of the bladder tumor on MRI was consistent with that of conventional malignant melanoma in case 1. Although no reports described the specific appearance of metastasis to genitourinary systems, appearance in the current cases was similar to that of metastasis to other organs. MRI may be useful in developing a differential diagnosis for bladder metastasis because common urothelial carcinoma revealed iso-intensity on T1WI and iso- to slightly high-intensity compared with the muscle layer on T2WI [ 17 , 18 ].

As follow-up, imaging examination at every 3–12 months for 2 years, then every 6–12 months for another 3 years was recommended for patients with stage IV NED with cutaneous melanoma [ 19 ]. Screening for metastasis was performed by CT scan at every 6 months in case 1, and bladder metastasis manifested as small tumors. Postoperative (after nephroureterectomy) follow-up was insufficient in case 2, and larger bladder melanoma was found, suggesting the significance of sufficient follow-up by routine imaging examination.

Metastatic malignant melanoma to the urinary tract has been reported as a rare secondary tumor. We successfully performed metastasectomy for bladder and ureteral metastases without local recurrence. In addition, long recurrence-free survival was observed in case 2. Complete resection for metastasis of malignant melanoma may have significant potential to improve survival.

Availability of data and materials

The supporting data and materials for this report are available on request from the corresponding author.

Abbreviations

Computed tomography

Magnetic resonance imaging

Transurethral resection of bladder tumor

No evidence of disease

Fink W, Zimpfer A, Ugurel S. Mucosal metastases in malignant melanoma. Oncology. 2003;26:249–51.

CAS   Google Scholar  

Stein BS, Kendall AR. Malignant melanoma of the genitourinary tract. J Urol. 1984;132:859–68.

Article   CAS   PubMed   Google Scholar  

Lee CS, Komenaka IK, Hurst-Wicker KS, et al . Management of metastatic malignant melanoma of the bladder. Urology. 2003;62:351.

PubMed   Google Scholar  

Patil RV, Woldu SL, Lucas E, et al . Metastatic melanoma to the bladder: case report and Review of the Literature. Urol Case Rep. 2017;10(11):33–6.

Article   Google Scholar  

Ogunremi O, Sirjuesingh D, Deshpande A. Metastatic melanoma to the urinary bladder: a rare cause of visible haematuria. Case Rep Urol. 2024;1:5516547.

Google Scholar  

Bates AW, Baithun SI. Secondary neoplasms of the bladder are histological mimics of nontransitional cell primary tumours: clinicopathological and histological features of 282 cases. Histopathology. 2000;36(1):32–40.

Velcheti V, Govindan R. Metastatic cancer involving bladder. A review Can J Urol. 2007;1:3443–8.

Goldstein AG. Metastatic carcinoma to the bladder. J Urol. 1967;98:209–15.

Sheehan EE, Greenberg SD, Scott R. Metastatic neoplasms of the bladder. J Urol. 1963;90:281–4.

Bates AW, Baithun SI. The significance of secondary neoplasms of the urinary and male genital tract. Virchows Arch. 2002;440(6):640–7.

Stoykov B, Velev D, Ali A, et al . Adrenal and ureteral metastasis of malignant melanoma: a case report and review of the literature. Urol Case Rep. 2022;45: 102286.

Article   PubMed   PubMed Central   Google Scholar  

DeSanto M, Strait R, Zopp J, et al . Metastatic melanoma to the proximal ureter. Urol Case Rep. 2023;50: 102543.

Macneil J, Hossack T. A case of metastatic melanoma in the ureter. Case Rep Urol. 2016. https://doi.org/10.1155/2016/1853015 .

Nair BC, Williams NC, Cui C, et al . Conjunctival melanoma: bladder and upper urinary tract metastases. J Clin Oncol. 2011;29(9):e216–9.

Article   PubMed   Google Scholar  

Deutsch GB, Flaherty DC, Kirchoff DD, et al . Association of surgical treatment, systemic therapy, and survival in patients with abdominal visceral melanoma metastases, 1965–2014: relevance of surgical cure in the era of modern systemic therapy. JAMA Surg. 2017;152:672–8.

Barat M, Guegan-Bart S, Cottereau AS, et al . CT, MRI and PET/CT features of abdominal manifestations of cutaneous melanoma: a review of current concepts in the era of tumor-specific therapies. Abdom Radiol (NY). 2021;46(5):2219–35.

Patnana M, Bronstein Y, Szklaruk J, et al . Multimethod imaging, staging, and spectrum of manifestations of metastatic melanoma. Clin Radiol. 2011;66(3):224–36.

Caglic I, Panebianco V, Vargas HA, et al . MRI of bladder cancer: local and nodal staging. J Magn Reson Imaging. 2020;52(3):649–67.

Swetter SM, Thompson JA, Albertini MR, et al . NCCN guidelines insights: melanoma: cutaneous, version 2.2021. J Natl Compr Canc Netw. 2021;19(4):364–76.

Download references

Acknowledgements

The authors thank Ms. Miyuki Akino of the Department of Urology, Faculty of Medicine, University of Miyazaki, for her assistance with ethical protocols.

We did not receive financial support for this study.

Author information

Authors and affiliations.

Department of Urology, Faculty of Medicine, Miyazaki University, 5200 Kihara, Kiyotake, Miyazaki, 889-1692, Japan

Takashi Ueno, Kaoru Ito, Takaya Murashima, Masato Fujii, Takahiro Nagai, Shoichiro Mukai, Atsuro Sawada & Toshiyuki Kamoto

Department of Pathology, Faculty of Medicine, Miyazaki University, Miyazaki, Japan

Takumi Kiwaki

Department of Urology, Dokkyo Medical University, Tochigi, Japan

Hironori Betsunoh

You can also search for this author in PubMed   Google Scholar

Contributions

TU drafted the manuscript, performed the examination, observation, and approved the final version of the manuscript. HB, KI, TM, MF, and TN performed examinations, surgery, and cared for the patient and approved the final version of the manuscript. TK (Takumi Kiwaki) diagnosed and reviewed the pathological specimens and approved the final version of the manuscript. SM, AS, and TK (Toshiyuki Kamoto) drafted the report and contributed the final version of the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Shoichiro Mukai .

Ethics declarations

Ethics approval and consent to participate.

This case report was approved by the Ethics Committee of Miyazaki University (approval number: C-0172). Consent to participant was obtained from the patient.

Consent for publication

Written informed consent was obtained from the patient for publication of this case report and any accompanying images. A copy of the written consent is available for review by the Editor-in Chief of this journal.

Competing interests

The authors declare that they have no conflict of interest.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Cite this article.

Ueno, T., Kiwaki, T., Betsunoh, H. et al. Metastasis of malignant melanoma to urinary tract: a case report. J Med Case Reports 18 , 396 (2024). https://doi.org/10.1186/s13256-024-04716-8

Download citation

Received : 10 May 2024

Accepted : 22 July 2024

Published : 28 August 2024

DOI : https://doi.org/10.1186/s13256-024-04716-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Urinary bladder
• Metastasectomy

Journal of Medical Case Reports

ISSN: 1752-1947

• Submission enquiries: Access here and click Contact Us
• General enquiries: [email protected]

A literature review of magnetic resonance imaging sequence advancements in visualizing functional neurosurgery targets

• Download PDF

Historically, preoperative planning for functional neurosurgery has depended on the indirect localization of target brain structures using visible anatomical landmarks. However, recent technological advances in neuroimaging have permitted marked improvements in MRI-based direct target visualization, allowing for refinement of “first-pass” targeting. The authors reviewed studies relating to direct MRI visualization of the most common functional neurosurgery targets (subthalamic nucleus, globus pallidus, and thalamus) and summarize sequence specifications for the various approaches described in this literature.

The peer-reviewed literature on MRI visualization of the subthalamic nucleus, globus pallidus, and thalamus was obtained by searching MEDLINE. Publications examining direct MRI visualization of these deep brain stimulation targets were included for review.

A variety of specialized sequences and postprocessing methods for enhanced MRI visualization are in current use. These include susceptibility-based techniques such as quantitative susceptibility mapping, which exploit the amount of tissue iron in target structures, and white matter attenuated inversion recovery, which suppresses the signal from white matter to improve the distinction between gray matter nuclei. However, evidence confirming the superiority of these sequences over indirect targeting with respect to clinical outcome is sparse. Future targeting may utilize information about functional and structural networks, necessitating the use of resting-state functional MRI and diffusion-weighted imaging.

• CONCLUSIONS

Specialized MRI sequences have enabled considerable improvement in the visualization of common deep brain stimulation targets. With further validation of their ability to improve clinical outcomes and advances in imaging techniques, direct visualization of targets may play an increasingly important role in preoperative planning.

• ABBREVIATIONS

The authors reviewed MRI sequences used for preoperative deep brain stimulation target visualization. A variety of MRI sequences have been designed for this purpose, each with their specific strengths and limitations. The authors provide a summary framework to optimize MRI visualization of deep brain stimulation targets.

F unctional neurosurgery is dedicated to modulating aberrant circuits associated with a wide range of neurological conditions. 1 Broadly speaking, this can be achieved by lesioning (e.g., radiosurgery, radiofrequency ablation, and MR-guided focused ultrasound) or electrical stimulation of key brain structures (deep brain stimulation [DBS]). While principally employed for the treatment of movement disorders such as Parkinson’s disease (PD), essential tremor (ET), and dystonia, the field, mainly driven by DBS, has seen its spectrum of potential indications expand to psychiatric (e.g., obsessive-compulsive disorder, Tourette syndrome, depression, and anorexia nervosa) and cognitive (e.g., Alzheimer’s disease) disorders. 2 While direct MRI visualization of the targeted brain structures was used early on, 3 generally it has been insufficient for preoperative planning. 4

Indirect targeting methods, which estimate the location of targets in relation to fixed and identifiable anatomical landmarks on MRI, have traditionally been used because DBS targets could not be visualized on ventriculography and CT. 4 However, indirect targeting fails to account for interindividual variability in the location of target structures. 5 To improve DBS targeting accuracy, indirect targeting methods were coupled with a variety of other techniques, such as intraoperative microelectrode recordings and intraoperative stimulation testing in awake patients. 6 However, these methods are associated with prolonged procedural times and require multiple penetrations of deep brain structures, increasing the risk of intra- and postoperative complications. 7

Routine brain MRI sequences acquired with standard field strengths and acquisition parameters have shortcomings in visualizing DBS targets. 8–10 However, with advances in stereotactic frames, MRI hardware, and pulse sequences, direct visualization of certain structures, such as the subthalamic nucleus (STN), is replacing indirect targeting methods for preoperative planning at some institutions. 7 , 11 Nonetheless, other structures, including thalamic nuclei, still require indirect targeting, as their direct radiological visualization remains challenging. 12 , 13 MRI techniques optimizing white/gray matter contrast 14 , 15 or leveraging differences in tissue composition such as iron content 16 have been developed to improve the delineation of DBS targets. The therapeutic effects achieved with DBS surgery hinge on the precise and selective modulation of the intended target structure, maximizing treatment efficacy while minimizing any off-target spillover into neighboring structures that might produce adverse effects. Therefore, it seems plausible that direct MRI targeting will be increasingly incorporated into preoperative planning at most institutions.

Here, the goal was to review the many different MRI techniques that have been developed to date to enhance visualization of the most common gray matter nuclei targeted with DBS while also discussing the relevant anatomy and clinical indications of these structures. Finally, we discuss the potential implications of expected MRI advancements on DBS surgery.

• Search Methods

A comprehensive search was conducted on June 1, 2020, using the MEDLINE database. The goal was to perform a scoping study, reviewing the literature to examine the extent of research activity, to summarize research findings, and to identify research gaps. The protocol for this scoping study was based on the framework proposed by Arksey and O’Malley, 17 with the incorporation of modifications proposed by Levac et al. 18 The search strategy employed terms related to “magnetic resonance imaging” and “deep brain stimulation” and the most common neurosurgical targets (see Appendix and Supplementary Fig. 1 for detailed search methods, syntaxes, and a protocol flow diagram). To maintain the clinical relevance of this review, only studies using clinically used field strengths (i.e., 1.5T or 3T MRI) were included.

Most Common DBS Targets

Figure 1 provides a visual comparison between nonoptimized routine (red outline) and optimized (green outline) MRI sequences from the literature to visualize DBS targets. Accompanying acquisition parameters are detailed in Table 1 . These optimized MR images were selected based on general trends among studies comparing MRI sequences for visualizing common DBS targets ( Table 2 ).

Examples of optimized sequences and postprocessing methods for enhanced MRI visualization of DBS targets. Zoomed-out and zoomed-in MR images of the STN ( A–F ), GPi ( G–L ), and thalamus ( M–R ) are shown in addition to the corresponding slice from the atlas of Mai et al. 98 The green outline identifies MR images from the literature aiming at improving visualization of DBS targets, whereas the red outline identifies routinely acquired (nonoptimized) T1W ( left ), T2W ( middle ), and PDW ( right ) images for comparison. Details pertaining to their publication and acquisition parameters are included in Table 1 . The STN is shown on coronal images, whereas the GPi and thalamus are shown on axial images. AMd = nucleus anteromedial; DL = nucleus dorsolateral; Inl = nucleus intermediolateral; Med = nucleus medial; PLR = prelemniscal radiations; Pu = putamen; SNC = substantia nigra compacta; Thal = thalamus; TSE = turbo spin echo; ZI = zona incerta. The atlas images were published in Atlas of the Human Brain (4th edition), Mai J, Majtanik M, Paxinos G, pp 239 and 406, Elsevier Academic Press, copyright Elsevier 2015. Panel C: reprinted with permission from Senova S, Hosomi K, Gurruchaga JM, et al. Three-dimensional SPACE fluid-attenuated inversion recovery at 3 T to improve subthalamic nucleus lead placement for deep brain stimulation in Parkinson’s disease: from preclinical to clinical studies. J Neurosurg. 2016;125(2):472–480. Panel D: reprinted from Neuroimage. 47(suppl 2), Sudhyadhom A, Haq IU, Foote KD, Okun MS, Bova FJ. A high resolution and high contrast MRI for differentiation of subcortical structures for DBS targeting: the Fast Gray Matter Acquisition T1 Inversion Recovery (FGATIR), pp T44–T52, copyright 2009, with permission from Elsevier. Panel E: reprinted by permission from Springer Nature. Int J Comput Assist Radiol Surg. 10(3):329–341, Multicontrast unbiased MRI atlas of a Parkinson’s disease population, Xiao Y, Fonov V, Beriault S, et al., copyright 2015, https://www.springer.com/journal/11548 . Panel F: reprinted with permission from Rasouli J, Ramdhani R, Panov FE, et al. Utilization of quantitative susceptibility mapping for direct targeting of the subthalamic nucleus during deep brain stimulation surgery. Oper Neurosurg (Hagerstown). 2018;14(4):412–419, by permission of Oxford University Press on behalf of the Congress of Neurological Surgeons. Panel J: Slightly modified with permission from Hirabayashi H, Tengvar M, Hariz MI. Stereotactic imaging of the pallidal target. Mov Disord. 17(suppl 3):S130–S134, Wiley & Sons, © 2002 Movement Disorder Society. Panel K: Slightly modified from Nowacki A, Fiechter M, Fichtner J, et al. Using MDEFT MRI sequences to target the GPi in DBS surgery. PLoS One. 2015;10(9):e0137868, © 2015 Nowicki et al. CC BY 4.0 International license ( https://creativecommons.org/licenses/by/4.0/ ). Panel L: Slightly modified with permission from Magn Reson Imaging. 63, Beaumont J, Saint-Jalmes H, Acosta O, et al. Multi T1-weighted contrast MRI with fluid and white matter suppression at 1.5T, pp 217–225, Crown Copyright © 2019. Published by Elsevier Inc. All rights reserved. Reprinted with permission. Panel P: reprinted with permission from Buentjen L, Kopitzki K, Schmitt FC, et al. Direct targeting of the thalamic anteroventral nucleus for deep brain stimulation by T1-weighted magnetic resonance imaging at 3 T. Stereotact Funct Neurosurg. 2014;92(1):25–30, © 2013 S. Karger AG, Basel. Panel Q: reprinted by permission from Springer Nature. Clin Neuroradiol . 27(4):511–518, Optimized depiction of thalamic substructures with a combination of T1-MPRAGE and phase: MPRAGE, Bender B, Wagner S, Klose U, copyright 2017. https://www.springer.com/journal/62/ . Panel R: reprinted from Brain Stimul. 5(4), Vassal F, Coste J, Derost P, et al. Direct stereotactic targeting of the ventrointermediate nucleus of the thalamus based on anatomical 1.5-T MRI mapping with a white matter attenuated inversion recovery (WAIR) sequence, pp 625–633, copyright 2012, with permission from Elsevier. Figure is available in color online only.

Optimized MRI acquisition parameters

Dyst = dystonia; FOV = field of view; NA = not applicable; pts = patients; Thal = thalamus; TSE = turbo spin echo.

Acquisition parameters used to optimize visualization of the STN, GPi, and thalamus as shown in Fig. 1 . Sequences in boldface type are routine (nonoptimized) sequences for comparison.

Comparison of MRI sequences for visualizing common DBS targets

C contour = contrast of the contour; CR = contrast ratio; DESPOT1 = driven equilibrium single-pulse observation of T1; FFE = fast field echo; FSE = fast spin echo; FSPGR = fast spoiled gradient echo; FSTIR = T1-weighted fast spin echo–based inversion recovery; HB = high bandwidth; HR = high resolution; MIP = minimum intensity projection; PADRE = phase difference enhanced imaging; R2* = R2* mapping; rater = qualitative scoring by raters; SE = spin echo;SPGR = spoiled gradient echo; SWAN = T2*-weighted angiography; TC = tissue contrast; TE40 = echo time 40.

Studies comparing MRI sequences to optimally visualize the STN, GPi, and thalamus are listed. Sequences in boldface type show the reference or “standard” sequence used in each study. The metric of image quality that the authors used for comparison is provided.

• Subthalamic Nucleus

Relevant Anatomy

The STN is a small (approximately 8 mm in the maximal transverse histological dimension 19 ), almond-shaped gray matter structure located inferior to the thalamus. It features complex neuroanatomical relationships, being bounded by the internal capsule anterolaterally, the substantia nigra (SN) ventrolaterally, cerebellothalamic fibers posteromedially, and fields of Forel and zona incerta superiorly ( Fig. 2A ). 20 , 21 The STN has three main functional subdivisions: a superior, posterior, and lateral sensorimotor area; a central associative area; and an emotive medial, anterior, and inferior tip. 21–23

Three-dimensional representations of DBS targets and relevant neighboring anatomy. A : The STN ( orange ) is shown medial to the internal capsule, lateral to the red nucleus, superior to the SN, and inferior to the thalamus. B : The GPi ( green ) is shown medial to the GPe, inferolateral to the thalamus, and superior to the optic tract. C : The thalamus ( pink ) is shown medial to the internal capsule and superior to the STN and hypothalamus. Structures are overlaid on coronal (A and C) and left oblique (B) T1W MR images of the brain (ICBM 2009b nonlinear asymmetrical Montreal Neurological Institute template). Anatomical structures are derived from the DISTAL atlas 50 and visualized in 3D with Lead-DBS ( www.lead-dbs.org ). Figure is available in color online only.

Clinical Indications for DBS Targeting

The sensorimotor STN is the main target for PD, whereas DBS of the associative and emotive STN has been investigated as a treatment for obsessive-compulsive disorder. 1 , 2

Direct MRI Visualization

T2-weighted (T2W) and inversion recovery (IR) imaging have classically been the most common approaches used to directly visualize the STN. 24 , 25 More recently, susceptibility-weighted imaging (SWI) and T2-star-weighted (T2*W) imaging have been employed. Finally, novel image processing techniques, such as quantitative susceptibility mapping (QSM) applied to SWI-based acquisitions, have shown promise in enhancing MRI visualization of the STN. 8 The STN is most reliably demarcated from the adjacent zona incerta and SN on coronal slices. 26

Most commonly, T2W sequences have been used for direct targeting of the STN. 25 In these sequences, the nucleus can be identified as a hypointense lentiform structure—presumably due to iron deposition 27 —measuring approximately 7 mm in the maximal radiological dimension. 19 The interface between the STN and SN is not always visible on T2W images, especially at 1.5T 28 (and also at 3T 29 ). Moreover, visualizing the STN on T2W sequences will only lead to improved targeting if stereotactic images are optimized for contrast and if they are processed to minimize geometric distortion. 30

IR sequences aim to enhance the visualization of a given structure by selectively suppressing certain tissues with a specific composition. When using a FLAIR sequence, which nulls the signal from fluid, the STN remains hypointense with reduced geometric distortion compared with routine T2W imaging. Although there is limited evidence comparing FLAIR sequences to other visualization techniques, Senova et al. 31 showed that preoperative targeting in PD patients with a 3T FLAIR sequence (3D SPACE [sampling perfection with application-optimized contrasts by using different flip angle evolution] FLAIR) was associated with both minimal geometric distortion and significantly higher contrast with surrounding structures, as well as better clinical outcomes at 12 months over routine T2W imaging ( Table 1 , Fig. 1C , and Supplementary Fig. 2C ). However, similar to T2W sequences, the STN borders adjacent to the SN remain difficult to delineate, even at 3T. 26

Other less commonly used IR sequences have also been used to visualize STN for DBS surgical planning; these include short T1 inversion recovery (STIR), which nulls signal from fat; phase-sensitive inversion recovery (PSIR); and more recently, fast gray matter acquisition T1 inversion recovery (FGATIR), intended to null the white matter signal. 25 Notably, PSIR is the only sequence in which the geometric distortion with a stereotactic head frame has been shown to be less than 1% at 1.5T, 32 while the STIR sequence has demonstrated increased contrast between the STN and SN at 3T, offering improved delineation of the inferior STN border. 33 Finally, FGATIR has shown promise in visualizing all STN borders in PD and ET patients, owing to increased contrast-to-noise ratio (CNR) ( Table 1 , Fig. 1D , and Supplementary Fig. 2D ). 34 Despite encouraging results, the use of these IR sequences in clinical settings remains relatively low to date, perhaps due to the specialized knowledge base required, single-vendor implementation, and the need for replication of relevant findings in larger studies.

SWI uses gradient echo (GRE) sequences, which enhance the effect created by magnetic susceptibility differences between tissues. In particular, these are valuable for imaging the increased iron content of the STN in the context of neurodegenerative diseases and aging. 35 The paramagnetic property of the STN can be leveraged to enhance its differentiation from neighboring structures. SWI images, as well as accompanying T2*W 26 , 36 and SWPI (susceptibility-weighted phase imaging), 37 have been successfully used to visualize all STN boundaries ( Table 1 , Fig. 1E , and Supplementary Fig. 2E ). However, these techniques are limited by geometric distortion, which has been shown to be as much as 0.8, 0.5, and 0.7 mm in the x, y, and z planes, respectively, in a fast low-angle shot (FLASH) sequence. 38 These distortions arise from the nonlocal susceptibility effect, which causes geometric distortion and, consequently, blurring and enlargement of STN borders, commonly observed in GRE sequences. 25 This occurs because the high iron content of the STN creates a local magnetic field in MRI, which induces the relaxation of protons in surrounding tissues, thereby producing a susceptibility effect outside the STN even in the absence of a susceptibility source. Moreover, the nonuniform distribution of iron in the STN disproportionately exaggerates the distortion of certain borders. 25

QSM is a postprocessing technique that allows for quantification and correction of geometric distortions when visualizing the STN with GRE sequences ( Table 1 , Fig. 1F , and Supplementary Fig. 2F ). 8 , 29 This technique reduces the nonlocal susceptibility effect by providing a clearer picture of tissue susceptibility and magnetic properties, irrespective of patient position (and thus STN orientation). 25 , 39 In addition, it provides a more accurate measurement of brain iron concentration, allowing for improved discrimination of surrounding iron-rich gray matter structures, including the SN, in PD patients. As with other SWI-based sequences, the geometric accuracy of QSM postprocessing has had limited validation in larger clinical studies, although Rasouli et al. showed a strong correlation of QSM with intraoperative microelectrode recording delineation of the STN in 25 PD patients. 29 Furthermore, the technique remains difficult to implement at most clinical centers, as image generation is technically demanding and often requires a significant amount of processing time. 40 However, online reconstruction techniques have shown promise in addressing these practical limitations, reducing the image construction time to less than 30 seconds on standard computers. 41 Nonetheless, QSM reconstruction algorithms remain a work in progress. 42

Table 2 lists studies that have compared sequences for direct STN visualization. In general, susceptibility-based sequences 16 , 26 , 43 – 45 and optimized IR sequences such as FGATIR and STIR 31 , 33 , 34 , 46 – 48 demonstrated superior signal-to-noise ratio (SNR) and CNR compared with the more traditionally used T2W and IR (e.g., FLAIR) sequences. Indeed, routine T2W and IR sequences have repeatedly offered suboptimal visualization of all STN borders at 1.5T. Unsurprisingly, higher-field-strength MRI (i.e., 3T) can improve STN border visualization with these sequences. 31 , 49

• Globus Pallidus

Named after its pallid appearance on anatomical specimens, the globus pallidus (GP) is a lens-shaped gray matter structure situated between the putamen and internal capsule ( Fig. 2B ). The putamen and GP, which together form the lentiform nucleus, are demarcated by the external medullary lamina. The GP itself is divided into two constituent parts by the medial medullary lamina: the GP internus (GPi) and GP externus (GPe). 34 The GPi borders the optic tract ventrally and the internal capsule medially. The motor component is functionally segregated in the posterior GPi. 50

After the STN, the GPi is the most common target for DBS in the management of PD. 2 Although both sites arguably provide similar motor benefits, the STN contributes to medication intake reduction, whereas the GPi may be better suited for PD patients with cognitive impairment and medication-associated dyskinesias. 51 , 52 The GPi is also the main target for dystonia and has shown promise in the treatment of Tourette syndrome. 2 While uncommonly used, stimulation of the GPe has also been shown to improve PD symptoms. 53

In our center’s experience, T2W and proton density–weighted (PDW) 15 ( Table 1 , Fig. 1J , and Supplementary Fig. 2J ) sequences are most commonly used for direct visualization of the GP. In contrast to the STN, the GPi is better appreciated on axial slices. 9 On T2W images, the GP can be seen as a hypointense structure, 54 whereas it is mildly hyperintense on PDW images. 15 At lower field strengths (i.e., 1.5T), these sequences generally visualize the optic tract, external medullary lamina and adjacent putamen, and the internal capsule bordering the posteromedial side of the GP. Delineation of additional boundaries, such as the medial medullary lamina, may not always be reliably obtained. 15 Among other sequences, Nowacki et al. 55 investigated the use of a T1-weighted (T1W) sequence in dystonia patients, specifically the modified equilibrium Fourier transform (MDEFT) technique, which is employed at high field strengths due to its advantageous contrast characteristics. Using this MDEFT approach at 3T field strength, the caudate putamen and pallidal subdivisions, the GPe and GPi, were well demarcated in most patients ( Table 1 and Fig. 1K ). Because the central trajectory was used in 88% of all cases, MDEFT-based planning was deemed accurate and reliable.

IR sequences, on which the pallidum appears as a hypointense structure, have also been used. IR spin echo sequences (e.g., IR-FSE [fast spin echo]) at 1.5T have been shown to visualize the optic tract and external medullary lamina. 54 FGATIR additionally allowed delineation of the internal medullary lamina. 34 FGATIR has further been modified to enhance the distinction between the GPi and GPe by using parameters suppressing the fluid and white matter sequence (FLAWS) ( Table 1 , Fig. 1L , and Supplementary Fig. 2L ). 56 In this study, FLAWS was generated through the registration of two contrasts, the standard T1W anatomical contrast of the brain (i.e., magnetization-prepared rapid acquisition with gradient echo [MPRAGE]) and suppression of the white matter signal (i.e., FGATIR), demonstrating enhanced visualization of subcortical structures in healthy participants.

As with the STN, susceptibility-based sequences permit direct visualization of the pallidum. T2*W and QSM sequences have been shown to discern the GPi and GPe in PD patients. 57 However, with an SWI-like sequence at 3T, Ide et al. 58 showed that the medial medullary lamina was less readily identifiable with increasing age, which may be related to increased mineralization in the GP and/or a loss of myelin.

Few studies have compared sequences for direct visualization of the GP and its subdivisions in a head-to-head manner ( Table 2 ). A handful of reports found that the GPi was best visualized using susceptibility-based sequences when compared with T1W, T2W, or IR sequences at 1.5T and 3T. 9 , 16 , 45 , 58 Another report found that, at 3T, the internal medullary lamina in PD and ET patients was better visualized on an FGATIR sequence compared with the more commonly employed FLAIR and T1W imaging (i.e., MPRAGE). 34 While these findings are not necessarily conflicting, additional studies comparing sequences would be helpful in establishing a consensus for optimal visualization of the pallidum and its internal architecture.

A large gray matter structure, the thalamus is located immediately above the hypothalamus and medial to the posterior limb of the internal capsule, forming the lateral wall of the third ventricle and floor of the lateral ventricles ( Fig. 2C ). Within the thalamus, the internal medullary lamina divides the structure into anterior, mediodorsal, ventral, and lateral groups, with each group comprising several distinct nuclei. 59 Functionally, the thalamus has a distinct topographical organization. In simple terms, the posterior part contributes to sensory processing, whereas the motor thalamic relay is located in the ventrolateral part. Finally, the anterior and mediodorsal nuclear groups are considered to be involved in limbic and associative functions, respectively. 59

Stimulation of the thalamic ventral intermediate nucleus (VIM) is well established for the management of ET and, to a lesser extent, tremor secondary to other pathological conditions, such as multiple sclerosis or stroke. 2 Modulation of the ventrocaudal (VC) nucleus has been performed to treat chronic pain disorders, particularly central poststroke pain. 60 Additionally, the centromedian nucleus (part of the intralaminar nuclei) has been targeted for multiple neurological and psychiatric indications, including Tourette syndrome, PD, and epilepsy, while stimulation of the anterior nucleus of the thalamus (ANT) has shown potential in suppressing global seizure activity in epilepsy patients. 1 , 2

Thalamic nuclei, including the VIM, are notoriously difficult to visualize on routine MRI sequences and often necessitate the use of atlas-derived coordinates for preoperative planning. 61 On routinely acquired T1W and T2W sequences, the thalamus is mildly hyperintense and hypointense, respectively. 62 Despite its many nuclei, it appears fairly homogeneous with little distinction between subdivisions. However, studies in healthy participants have shown that the inversion time of T1W sequences may be optimized, allowing suppression of gray matter. 14 , 63 The resulting gray and white matter differentiation enables identification of the main thalamic groups: anterior, dorsomedial, lateral, and ventral. 14 Optimizing the repetition time of T1W imaging (i.e., MPRAGE) has also been shown to enable specific delineation of the ANT, improving targeting prior to DBS epilepsy surgery ( Table 1 , Fig. 1P , and Supplementary Fig. 2P ). 63 To further optimize visualization of the thalamic nuclei, MPRAGE has been combined with phase data from 3D GRE sequences, which enabled the distinction of additional thalamic substructures such as the VIM ( Table 1 and Fig. 1Q ). 64 These techniques have yet to be demonstrated in diseased populations.

In one study using a 3T PDW sequence, it was possible to visualize the VIM in healthy subjects as a mildly hypointense band crossing the anterior third of the thalamus, from lateral to medial. 65 However, it was inconsistently seen at 1.5T. Furthermore, the sensory thalamic nuclei (i.e., VC nucleus) was seen as another hypointense band located posteriorly. 65 Using a PDW sequence at 3T, direct targeting of the VIM has been successfully performed in a tremor patient. 66

IR sequences have also enabled visualization of the VIM. Specifically, studies have shown the VIM to be slightly hyperintense relative to the posterior nuclei on STIR sequences. 67 , 68 IR sequences, including STIR and FGATIR, have also been used for targeting of the ANT based on delineation of the mammillothalamic tract, which terminates in the ANT. 62 , 69 , 70 An IR sequence suppressing signal from white matter (i.e., white matter attenuated inversion recovery [WAIR]) has demonstrated significant enhancement of contrast between different gray matter territories in PD and ET patients, with promise in visualizing the internal subdivisions of the thalamus ( Table 1 , Fig. 1R , and Supplementary Fig. 2R ). 13 On WAIR, the VIM appears as a hypointense band crossing the ventrolateral region of the thalamus relative to the surrounding nuclei.

Across the very small number of studies comparing thalamic visualization sequences, IR sequences have been found to be superior to routine T1W imaging ( Table 2 ). 62 , 69

• Limitations

As recently as 15 years ago, indirect targeting based on anatomical landmarks was the mainstay of preoperative surgical planning for most functional neurosurgery services. However, advances in MRI hardware and techniques have allowed direct targeting to become more accessible and clinically feasible. 25 Despite these improvements, there is limited consensus on the optimal MRI sequences for direct visualization of common DBS targets. While addressing this issue, this review contains significant limitations, highlighting gaps in the literature that future studies may seek to confront. First, a large proportion of the studies were performed in healthy volunteers, which may not accurately reflect the radiological findings in DBS patients ( Supplementary Table 1 ). For example, patients with PD, the most common DBS indication, demonstrate more pronounced brain atrophy and decreased white matter volume than healthy subjects. 71 , 72 This phenomenon, compounded by normal age-related atrophy, has been hypothesized to underlie the decreased STN visibility in older PD patients. 10 Second, demonstrable improvements in image quality and the use of novel sequences are limited by the persistent requirement of using specific head coils (usually lower SNR and potentially geometric distortion) that may not physically accommodate stereotactic head frames. Alternatively, frameless techniques or MRI/CT coregistration may be used. However, this may add errors of coregistration that reduce first-pass accuracy. 73 Third, an important practical consideration of these novel sequences is that while the delineation of DBS targets is improved, it may be difficult to visualize frame or frameless fiducials, as well as other anatomical landmarks, such as the anterior and posterior commissures, which are commonly used to provide the overall anatomical picture necessary for preoperative planning. Consequently, an MRI/MRI registration—between a sequence visualizing DBS targets and an anatomical sequence—may be required. The accuracy of coregistering these optimized sequences to anatomical MRI sequences has not been thoroughly assessed in the literature. For example, Rasouli et al. 29 described a registration method between QSM and T1W without providing accuracy measurement. Novel composite sequences simultaneously acquire a sequence providing both anatomical details and DBS target visualization, which may mitigate the problem. 36 Additionally, to leverage the advantages provided by these optimized sequences, generally 3T and coils incompatible with stereotactic head frames have to be used, which creates an additional step of coregistration with a CT scan, potentially introducing geometric error. While such coregistrations inherently introduce error, a systematic review of MRI/CT fusion for localization of electrode placement concluded that fusion was an accurate, reliable, and safe modality for assessing electrode location. 74 Fourth, studies that quantitatively compare sequences with regard to visualization and clinical outcomes remain few and far between, with little evidence establishing the superiority of one sequence over another. Most importantly, geometric distortions associated with most of these optimized sequences remain to be tested. Finally, studies also investigating whether direct target visualization definitively improves clinical outcomes when compared with indirect targeting should be done.

• Summary Framework to Optimize MRI Visualization of DBS Targets

As highlighted in Fig. 1 , marked improvements in direct visualization of targets when using optimized, rather than routine (nonoptimized), sequences have been made. Importantly, as discussed in this review, there is little quantitative evidence favoring specific sequences for visualization of each DBS target. In general, Table 2 seems to suggest that IR sequences, for example, FGATIR, provide improved visualization for the STN, GPi, and thalamus. This is a potential option for centers that do not have the requisite expertise needed to design their own optimized MRI protocols, or implement postprocessing techniques such as QSM, for each DBS target. Finally, when implementing new sequences in their surgical planning, neurosurgeons should audit their own targeting accuracy and assess for systematic errors that may have originated from geometric distortions and other sources of error. 75 , 76

• Future Directions

Direct visualization of the most common DBS targets, namely the STN, GPi, and thalamus, has markedly improved in recent years. Further improvement may be expected as ultra–high-field (UHF) MRI becomes more widely available ( Fig. 3A and B ). Higher magnetic field strengths offer increased SNR, which in turn allows increased spatial resolution, permitting the delineation of smaller neuroanatomical structures. 77 – 79 UHF MRI also confers a superior CNR, improving the ability to differentiate between two small abutting structures. 78 , 79 Given these advantages, it is unsurprising that UHF MRI has been shown to better visualize DBS targets than 1.5T and 3T with comparable acquisition times. 80 However, while higher magnetic field strengths may improve visualization per se, they are also more prone to susceptibility effects and image distortions, 25 , 81 theoretically leading to a greater risk of mistargeting. Furthermore, UHF MRI has not been utilized in conjunction with commercially available stereotactic frames to date. This necessitates that UHF MR images be coregistered with stereotactic images acquired using another modality (e.g., CT) for preoperative targeting, a step that can introduce registration errors. 73 Finally, the risks of DBS systems in UHF MRI have not been thoroughly evaluated, potentially limiting the widespread clinical applicability of this technology. 82 Taking these aforementioned caveats into consideration, it is clear that further studies are needed to compare UHF MRI with conventional MRI (1.5T and 3T) for surgical targeting, with clinical outcomes being used as the primary endpoint. With the advent of image distortion correction methods, 83 continued testing is required to elucidate potential benefits, obstacles, and tradeoffs presented by UHF MRI.

Potential implications of expected MRI advancements on DBS surgery. A : Coronal 7T white matter–nulled T1W MPRAGE image at the level of the thalamus from a healthy individual. Reprinted from Neuroimage 84, Tourdias T, Saranathan M, Levesque IR, Su J, Rutt BK, Visualization of intra-thalamic nuclei with optimized white matter–nulled MPRAGE at 7T, Neuroimage , pp 534–545, copyright 2014, with permission from Elsevier. B : Coronal 7T balanced steady-state free precession (bSSFP), which is a modified fast imaging employing steady-state acquisition (FIESTA) sequence, obtained at the level of the thalamus from a healthy individual. Reprinted from Lenglet C, Abosch A, Yacoub E, et al. Comprehensive in vivo mapping of the human basal ganglia and thalamic connectome in individuals using 7T MRI. PLoS One . 2012;7(1):e29153, copyright 2012 Lenglet et al. CC BY 3.0 license ( https://creativecommons.org/licenses/by/3.0/ ). C–E : White matter tracts derived from DWI of approximately 1000 healthy subjects overlaid on a sagittal 7T FLASH MR image. The dentato-rubro-thalamic tract, a target for tremor, is shown ( green ) alongside a 3D representation of the thalamus ( blue ) derived from the DISTAL atlas (C). 50 The medial forebrain bundle, which has been targeted for treatment of depression and obsessive-compulsive disorder, is shown ( purple ; D). The cingulum bundle ( red ), minor forceps ( yellow ), and uncinate fasciculus ( blue ), which have been used to guide targeting of the subcallosal region for depression, are shown (E). Panels C–E were constructed using data described in Edlow et al. 99 Figure is available in color online only.

The desire to expand the indications for DBS and improve on traditional targets has contributed to a paradigm shift in preoperative targeting. Rather than discrete structures, such as deep gray matter nuclei, optimal targets may include white matter pathways 84 , 85 or focal hubs of functional networks, 86 which are entities that are not necessarily appreciated on structural sequences at any field strength. Moreover, there is a growing appreciation that optimal targets may differ between patients, reflecting both heterogeneity within specific disorders and underlying interindividual differences in brain “wiring.” To be appreciated, white matter tracts and functional networks require both highly specialized MRI sequences—diffusion-weighted imaging (DWI) tractography and resting-state functional MRI (rsfMRI), respectively—and fairly complex postprocessing. While rsfMRI networks have been shown to predict improvement in STN stimulation for PD, this technique remains experimental. 86 Conversely, structural connectivity profiles have been shown to retrospectively correlate with clinical outcome, 61 and tractography-based targeting of the dentato-rubro-thalamic tract (DRTT) and medial forebrain bundle has already been employed as a clinical treatment for ET and psychiatric disorders, respectively ( Fig. 3C–E ). 87 , 88 DRTT has been seldom used to prospectively guide DBS targeting. 89 However, the protocol for a randomized controlled trial comparing the efficacy of VIM DBS (the traditional DBS target for tremor) and DRTT DBS has been published. 90 Outside of DBS, prospective DRTT targeting with MRI-guided focused ultrasound was shown to provide excellent symptom relief in patients with tremor. 91 Since many of the proposed targets trialed for psychiatric disorders, such as the medial forebrain bundle for treatment-resistant depression and obsessive-compulsive disorder, are white matter structures, tractography-based targeting is required. Tractography has also been used to functionally segment gray matter targets, such as the STN or thalamus, based on their white matter projections to the cortex, thereby potentially offering an alternative method to demarcate zones of clinical interest within these structures. 92 , 93 This method has not been thoroughly investigated in a prospective fashion. Overall, rsfMRI and DWI tractography provide an opportunity to both refine current targets using network-centered approaches and better visualize new or emerging targets that are not amenable to visualization with structural MRI sequences. The limitations of these emerging techniques need to be considered, particularly their validation in prospective studies, which is generally lacking.

• Conclusions

Due to technological advances in neuroimaging, most DBS targets can currently be visualized on MRI to some degree, providing an adjunct to indirect targeting. Progress in this field largely stems from the development of optimized sequences and acquisition parameters and has also been furthered by the increasing use of 3T MRI in clinical settings. It is expected that direct visualization will continue to improve, eventually enabling sufficient visualization of additional targets such as the pedunculopontine nucleus, 94 which are thus far difficult to appreciate.

While direct visualization of DBS targets has the benefit of taking into account interpatient anatomical variability and encouraging more individualized preoperative planning, studies are needed to definitely establish the superiority of direct targeting over indirect targeting, and to establish which visualization techniques have the highest spatial fidelity for each target. Upcoming developments in this field are likely to relate to UHF MRI, which is expected to provide markedly higher SNR and CNR, along with the emergence of techniques such as rsfMRI and DWI tractography. These advancements in MRI techniques offer the possibility of refining existing targets and discovering new targets by tapping into distributed functional or structural networks.

• Acknowledgments

We would like to acknowledge Asma Naheed and Nicole Bennett for providing their technical expertise to acquire the nonoptimized sequences included in Table 1 and Fig. 1 .

This work is supported by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG NE 2276/1-1) (C.N.), the RR Tasker Chair in Functional Neurosurgery at University Health Network, and a Tier 1 Canada Research Chair in Neuroscience. L.Z. is supported by the National Institute for Health Research University College London Hospitals Biomedical Research Centre. The corresponding author (A.M.L.) confirms that he had full access to all the data in the study and had final responsibility for the decision to submit for publication.

A comprehensive search was conducted on June 1, 2020, using the MEDLINE database. The goal was to perform a scoping study, 17 , 18 reviewing the literature to examine the extent of research activity, to summarize research findings, and to identify research gaps. Compared with systematic reviews, scoping studies are a type of literature review that tend to address broader topics that may include many different study designs. Consequently, they are less likely to address very specific research questions and assess the quality of included studies. Thus, a scoping study is one method for reviewing the literature and mapping key concepts. The limitations of this method include the lack of evidence, the lack of quality appraisal, and the disregard of the relative weight of evidence in favor of a particular conclusion. Nevertheless, we provide a rigorous and transparent method to summarize the field of MRI sequence advancements in visualizing functional neurosurgery targets. Consultation of key stakeholders, including neurosurgeons (M.P., L.Z., and S.K.K.), a neurologist (A.F.), neuroradiologists (W.K. and D.M.), and a physicist (C.J.S.), provided an opportunity for insights beyond those found in the literature.

Search terms consisted of a combination of exploded MeSH and free-text terms that comprised (exp Magnetic Resonance Imaging OR (MRI OR magnetic resonance imag*).kw,tw.) AND (exp Electric Stimulation Therapy OR (stimulat* OR DBS).kw,tw.) AND (exp limbic system, OR exp subthalamus, OR exp thalamus) OR (STN OR subthal* OR thalam* OR GPi OR GPe OR globus pallidus).kw,tw.)). In addition, 22 records identified through a preliminary search were added to the 3002 publications retrieved by the final search strategy. Using an online literature review management software (www.covidence.org), two reviewers (C.T.C. and A.T.) initially screened the titles and abstracts and subsequently screened full-text articles. Disagreements were settled by a consensus decision after discussion with a third reviewer (A.B.). After two rounds of screening, 62 eligible articles were compiled and extracted ( Supplementary Fig. 1 ). Study inclusion criteria were human studies that pertained to direct visualization or segmentation of the STN, GP, and thalamus using clinically used field strengths (i.e., 1.5T or 3T MRI). The exclusion criteria were lack of full text, review articles, and articles in languages other than English, German, or French. Data extraction was performed independently by two authors (C.T.C. and A.T.) using a preconstructed spreadsheet with the following headings: publication year, patient demographics, surgical target, and acquisition parameters. The findings were summarized in a narrative fashion.

• Disclosures

Dr. Zrinzo: consultant for Medtronic, Boston Scientific, and Elekta. Dr. Kalia: consultant for and honoraria from Medtronic. Dr. Fasano: grants, personal fees, and nonfinancial support from Abbvie, Medtronic, and Boston Scientific; personal fees from Sunovion, Chiesi Farmaceutici, and UCB; and grants and personal fees from Ipsen. Dr. Lozano: consultant for Medtronic, Boston Scientific, Abbott, and Insightec; and scientific director of Functional Neuromodulation.

• Author Contributions

Conception and design: Lozano, Boutet, Loh. Acquisition of data: Boutet, Loh, Chow, Taha. Analysis and interpretation of data: Lozano, Boutet, Loh, Chow, Taha, Elias, Neudorfer, Germann, Paff. Drafting the article: Boutet, Loh, Chow, Elias, Neudorfer, Germann, Paff. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Administrative/technical/material support: Lozano. Study supervision: Lozano.

Supplemental Information

• Online-Only Content

Supplemental material is available with the online version of the article.

Supplementary Table and Figures . https://thejns.org/doi/suppl/10.3171/2020.8.JNS201125 .

Lozano AM , Lipsman N . Probing and regulating dysfunctional circuits using deep brain stimulation . Neuron . 2013 ; 77 ( 3 ): 406 – 424 .

• Search Google Scholar
• Export Citation

Lozano AM , Lipsman N , Bergman H , et al. Deep brain stimulation: current challenges and future directions . Nat Rev Neurol . 2019 ; 15 ( 3 ): 148 – 160 .

Leksell L , Leksell D , Schwebel J . Stereotaxis and nuclear magnetic resonance . J Neurol Neurosurg Psychiatry . 1985 ; 48 ( 1 ): 14 – 18 .

Lemaire JJ , Coste J , Ouchchane L , et al. Brain mapping in stereotactic surgery: a brief overview from the probabilistic targeting to the patient-based anatomic mapping . Neuroimage . 2007 ; 37 ( suppl 1 ): S109 – S115 .

Patel NK , Khan S , Gill SS . Comparison of atlas- and magnetic-resonance-imaging-based stereotactic targeting of the subthalamic nucleus in the surgical treatment of Parkinson’s disease . Stereotact Funct Neurosurg . 2008 ; 86 ( 3 ): 153 – 161 .

Bejjani BP , Dormont D , Pidoux B , et al. Bilateral subthalamic stimulation for Parkinson’s disease by using three-dimensional stereotactic magnetic resonance imaging and electrophysiological guidance . J Neurosurg . 2000 ; 92 ( 4 ): 615 – 625 .

Zrinzo L , Hariz M , Hyam JA , et al. A paradigm shift toward MRI-guided and MRI-verified DBS surgery. Letter . J Neurosurg . 2016 ; 124 ( 4 ): 1135 – 1138 .

Boutet A , Gramer R , Steele CJ , et al. Neuroimaging technological advancements for targeting in functional neurosurgery . Curr Neurol Neurosci Rep . 2019 ; 19 ( 7 ): 42 .

Nölte IS , Gerigk L , Al-Zghloul M , et al. Visualization of the internal globus pallidus: sequence and orientation for deep brain stimulation using a standard installation protocol at 3.0 Tesla . Acta Neurochir (Wien) . 2012 ; 154 ( 3 ): 481 – 494 .

Ranjan M , Boutet A , Xu DS , et al. Subthalamic nucleus visualization on routine clinical preoperative MRI scans: a retrospective study of clinical and image characteristics predicting its visualization . Stereotact Funct Neurosurg . 2018 ; 96 ( 2 ): 120 – 126 .

Aviles-Olmos I , Kefalopoulou Z , Tripoliti E , et al. Long-term outcome of subthalamic nucleus deep brain stimulation for Parkinson’s disease using an MRI-guided and MRI-verified approach . J Neurol Neurosurg Psychiatry . 2014 ; 85 ( 12 ): 1419 – 1425 .

Benabid AL , Koudsie A , Benazzouz A , et al. Imaging of subthalamic nucleus and ventralis intermedius of the thalamus . Mov Disord . 2002 ; 17 ( suppl 3 ): S123 – S129 .

Vassal F , Coste J , Derost P , et al. Direct stereotactic targeting of the ventrointermediate nucleus of the thalamus based on anatomic 1.5-T MRI mapping with a white matter attenuated inversion recovery (WAIR) sequence . Brain Stimul . 2012 ; 5 ( 4 ): 625 – 633 .

Bender B , Mänz C , Korn A , et al. Optimized 3D magnetization-prepared rapid acquisition of gradient echo: identification of thalamus substructures at 3T . AJNR Am J Neuroradiol . 2011 ; 32 ( 11 ): 2110 – 2115 .

Hirabayashi H , Tengvar M , Hariz MI . Stereotactic imaging of the pallidal target . Mov Disord . 2002 ; 17 ( suppl 3 ): S130 – S134 .

Liu T , Eskreis-Winkler S , Schweitzer AD , et al. Improved subthalamic nucleus depiction with quantitative susceptibility mapping . Radiology . 2013 ; 269 ( 1 ): 216 – 223 .

Arksey H , O’Malley L . Scoping studies: towards a methodological framework . Int J Soc Res Methodol . 2005 ; 8 ( 1 ): 19 – 32 .

Levac D , Colquhoun H , O’Brien KK . Scoping studies: advancing the methodology . Implement Sci . 2010 ; 5 : 69 .

Mavridis I , Boviatsis E , Anagnostopoulou S . Anatomy of the human subthalamic nucleus: a combined morphometric study . Anat Res Int . 2013 ; 2013 : 319710 .

Weintraub DB , Zaghloul KA . The role of the subthalamic nucleus in cognition . Rev Neurosci . 2013 ; 24 ( 2 ): 125 – 138 .

Hamani C , Saint-Cyr JA , Fraser J , et al. The subthalamic nucleus in the context of movement disorders . Brain . 2004 ; 127 ( pt 1 ): 4 – 20 .

Accolla EA , Dukart J , Helms G , et al. Brain tissue properties differentiate between motor and limbic basal ganglia circuits . Hum Brain Mapp . 2014 ; 35 ( 10 ): 5083 – 5092 .

Lambert C , Zrinzo L , Nagy Z , et al. Confirmation of functional zones within the human subthalamic nucleus: patterns of connectivity and sub-parcellation using diffusion weighted imaging . Neuroimage . 2012 ; 60 ( 1 ): 83 – 94 .

Brunenberg EJ , Platel B , Hofman PA , et al. Magnetic resonance imaging techniques for visualization of the subthalamic nucleus . J Neurosurg . 2011 ; 115 ( 5 ): 971 – 984 .

Chandran AS , Bynevelt M , Lind CR . Magnetic resonance imaging of the subthalamic nucleus for deep brain stimulation . J Neurosurg . 2016 ; 124 ( 1 ): 96 – 105 .

Kerl HU , Gerigk L , Pechlivanis I , et al. The subthalamic nucleus at 3.0 Tesla: choice of optimal sequence and orientation for deep brain stimulation using a standard installation protocol . J Neurosurg . 2012 ; 117 ( 6 ): 1155 – 1165 .

Dormont D , Ricciardi KG , Tandé D , et al. Is the subthalamic nucleus hypointense on T2-weighted images? A correlation study using MR imaging and stereotactic atlas data . AJNR Am J Neuroradiol . 2004 ; 25 ( 9 ): 1516 – 1523 .

Danish SF , Jaggi JL , Moyer JT , et al. Conventional MRI is inadequate to delineate the relationship between the red nucleus and subthalamic nucleus in Parkinson’s disease . Stereotact Funct Neurosurg . 2006 ; 84 ( 1 ): 12 – 18 .

Rasouli J , Ramdhani R , Panov FE , et al. Utilization of quantitative susceptibility mapping for direct targeting of the subthalamic nucleus during deep brain stimulation surgery . Oper Neurosurg (Hagerstown) . 2018 ; 14 ( 4 ): 412 – 419 .

Zonenshayn M , Rezai AR , Mogilner AY , et al. Comparison of anatomic and neurophysiological methods for subthalamic nucleus targeting . Neurosurgery . 2000 ; 47 ( 2 ): 282 – 294 .

Senova S , Hosomi K , Gurruchaga JM , et al. Three-dimensional SPACE fluid-attenuated inversion recovery at 3 T to improve subthalamic nucleus lead placement for deep brain stimulation in Parkinson’s disease: from preclinical to clinical studies . J Neurosurg . 2016 ; 125 ( 2 ): 472 – 480 .

Ishimori T , Nakano S , Mori Y , et al. Preoperative identification of subthalamic nucleus for deep brain stimulation using three-dimensional phase sensitive inversion recovery technique . Magn Reson Med Sci . 2007 ; 6 ( 4 ): 225 – 229 .

Kitajima M , Korogi Y , Kakeda S , et al. Human subthalamic nucleus: evaluation with high-resolution MR imaging at 3.0 T . Neuroradiology . 2008 ; 50 ( 8 ): 675 – 681 .

Sudhyadhom A , Haq IU , Foote KD , et al. A high resolution and high contrast MRI for differentiation of subcortical structures for DBS targeting: the Fast Gray Matter Acquisition T1 Inversion Recovery (FGATIR) . Neuroimage . 2009 ; 47 ( suppl 2 ): T44 – T52 .

Brass SD , Chen NK , Mulkern RV , Bakshi R . Magnetic resonance imaging of iron deposition in neurological disorders . Top Magn Reson Imaging . 2006 ; 17 ( 1 ): 31 – 40 .

Xiao Y , Fonov V , Bériault S , et al. Multi-contrast unbiased MRI atlas of a Parkinson’s disease population . Int J CARS . 2015 ; 10 ( 3 ): 329 – 341 .

Vertinsky AT , Coenen VA , Lang DJ , et al. Localization of the subthalamic nucleus: optimization with susceptibility-weighted phase MR imaging . AJNR Am J Neuroradiol . 2009 ; 30 ( 9 ): 1717 – 1724 .

Watanabe Y , Lee CK , Gerbi BJ . Geometrical accuracy of a 3-tesla magnetic resonance imaging unit in Gamma Knife surgery . J Neurosurg . 2006 ; 105 ( suppl ): 190 – 193 .

Li J , Chang S , Liu T , et al. Reducing the object orientation dependence of susceptibility effects in gradient echo MRI through quantitative susceptibility mapping . Magn Reson Med . 2012 ; 68 ( 5 ): 1563 – 1569 .

Schweser F , Deistung A , Reichenbach JR . Foundations of MRI phase imaging and processing for Quantitative Susceptibility Mapping (QSM) . Z Med Phys . 2016 ; 26 ( 1 ): 6 – 34 .

Schweser F , Deistung A , Sommer K , Reichenbach JR . Toward online reconstruction of quantitative susceptibility maps: superfast dipole inversion . Magn Reson Med . 2013 ; 69 ( 6 ): 1582 – 1594 .

Langkammer C , Schweser F , Shmueli K , et al. Quantitative susceptibility mapping: report from the 2016 reconstruction challenge . Magn Reson Med . 2018 ; 79 ( 3 ): 1661 – 1673 .

Lefranc M , Derrey S , Merle P , et al. High-resolution 3-dimensional T2*-weighted angiography (HR 3-D SWAN): an optimized 3-T magnetic resonance imaging sequence for targeting the subthalamic nucleus . Neurosurgery . 2014 ; 74 ( 6 ): 615 – 627 .

Nagahama H , Suzuki K , Shonai T , et al. Comparison of magnetic resonance imaging sequences for depicting the subthalamic nucleus for deep brain stimulation . Radiol Phys Technol . 2015 ; 8 ( 1 ): 30 – 35 .

O’Gorman RL , Shmueli K , Ashkan K , et al. Optimal MRI methods for direct stereotactic targeting of the subthalamic nucleus and globus pallidus . Eur Radiol . 2011 ; 21 ( 1 ): 130 – 136 .

Heo YJ , Kim SJ , Kim HS , et al. Three-dimensional fluid-attenuated inversion recovery sequence for visualisation of subthalamic nucleus for deep brain stimulation in Parkinson’s disease . Neuroradiology . 2015 ; 57 ( 9 ): 929 – 935 .

Sarkar SN , Sarkar PR , Papavassiliou E . Subthalamic nuclear tissue contrast in inversion recovery MRI decreases with age in medically refractory Parkinson’s disease . J Neuroimaging . 2015 ; 25 ( 2 ): 303 – 306 .

Ben-Haim S , Gologorsky Y , Monahan A , et al. Fiducial registration with spoiled gradient-echo magnetic resonance imaging enhances the accuracy of subthalamic nucleus targeting . Neurosurgery . 2011 ; 69 ( 4 ): 870 – 875 .

van Laar PJ , Oterdoom DL , Ter Horst GJ , et al. Surgical accuracy of 3-tesla versus 7-tesla magnetic resonance imaging in deep brain stimulation for Parkinson disease . World Neurosurg . 2016 ; 93 : 410 – 412 .

Ewert S , Plettig P , Li N , et al. Toward defining deep brain stimulation targets in MNI space: a subcortical atlas based on multimodal MRI, histology and structural connectivity . Neuroimage . 2018 ; 170 : 271 – 282 .

Mansouri A , Taslimi S , Badhiwala JH , et al. Deep brain stimulation for Parkinson’s disease: meta-analysis of results of randomized trials at varying lengths of follow-up . J Neurosurg . 2018 ; 128 ( 4 ): 1199 – 1213 .

Odekerken VJ , van Laar T , Staal MJ , et al. Subthalamic nucleus versus globus pallidus bilateral deep brain stimulation for advanced Parkinson’s disease (NSTAPS study): a randomised controlled trial . Lancet Neurol . 2013 ; 12 ( 1 ): 37 – 44 .

Vitek JL , Hashimoto T , Peoples J , et al. Acute stimulation in the external segment of the globus pallidus improves parkinsonian motor signs . Mov Disord . 2004 ; 19 ( 8 ): 907 – 915 .

Starr PA , Vitek JL , DeLong M , Bakay RA . Magnetic resonance imaging-based stereotactic localization of the globus pallidus and subthalamic nucleus . Neurosurgery . 1999 ; 44 ( 2 ): 303 – 314 .

Nowacki A , Fiechter M , Fichtner J , et al. Using MDEFT MRI sequences to target the GPi in DBS surgery . PLoS One . 2015 ; 10 ( 9 ): e0137868 .

Beaumont J , Saint-Jalmes H , Acosta O , et al. Multi T1-weighted contrast MRI with fluid and white matter suppression at 1.5 T . Magn Reson Imaging . 2019 ; 63 : 217 – 225 .

Ide S , Kakeda S , Ueda I , et al. Internal structures of the globus pallidus in patients with Parkinson’s disease: evaluation with quantitative susceptibility mapping (QSM) . Eur Radiol . 2015 ; 25 ( 3 ): 710 – 718 .

Ide S , Kakeda S , Yoneda T , et al. Internal structures of the globus pallidus in patients with Parkinson’s disease: evaluation with phase difference-enhanced imaging . Magn Reson Med Sci . 2017 ; 16 ( 4 ): 304 – 310 .

Nieuwenhuys R , Voogd J , van Huijzen C . The Human Central Nervous System . 4th ed. Springer ; 2008 .

Hamani C , Schwalb JM , Rezai AR , et al. Deep brain stimulation for chronic neuropathic pain: long-term outcome and the incidence of insertional effect . Pain . 2006 ; 125 ( 1-2 ): 188 – 196 .

Akram H , Dayal V , Mahlknecht P , et al. Connectivity derived thalamic segmentation in deep brain stimulation for tremor . Neuroimage Clin . 2018 ; 18 : 130 – 142 .

Grewal SS , Middlebrooks EH , Kaufmann TJ , et al. Fast gray matter acquisition T1 inversion recovery MRI to delineate the mammillothalamic tract for preoperative direct targeting of the anterior nucleus of the thalamus for deep brain stimulation in epilepsy . Neurosurg Focus . 2018 ; 45 ( 2 ): E6 .

Buentjen L , Kopitzki K , Schmitt FC , et al. Direct targeting of the thalamic anteroventral nucleus for deep brain stimulation by T1-weighted magnetic resonance imaging at 3 T . Stereotact Funct Neurosurg . 2014 ; 92 ( 1 ): 25 – 30 .

Bender B , Wagner S , Klose U . Optimized depiction of thalamic substructures with a combination of T1-MPRAGE and phase: MPRAGE . Clin Neuroradiol . 2017 ; 27 ( 4 ): 511 – 518 .

Spiegelmann R , Nissim O , Daniels D , et al. Stereotactic targeting of the ventrointermediate nucleus of the thalamus by direct visualization with high-field MRI . Stereotact Funct Neurosurg . 2006 ; 84 ( 1 ): 19 – 23 .

Sidiropoulos C , Mubita L , Krstevska S , Schwalb JM . Successful Vim targeting for mixed essential and parkinsonian tremor using intraoperative MRI . J Neurol Sci . 2015 ; 358 ( 1-2 ): 488 – 489 .

Alterman RL , Reiter GT , Shils J , et al. Targeting for thalamic deep brain stimulator implantation without computer guidance: assessment of targeting accuracy . Stereotact Funct Neurosurg . 1999 ; 72 ( 2-4 ): 150 – 153 .

Yamada K , Akazawa K , Yuen S , et al. MR imaging of ventral thalamic nuclei . AJNR Am J Neuroradiol . 2010 ; 31 ( 4 ): 732 – 735 .

Jiltsova E , Möttönen T , Fahlström M , et al. Imaging of anterior nucleus of thalamus using 1.5T MRI for deep brain stimulation targeting in refractory epilepsy . Neuromodulation . 2016 ; 19 ( 8 ): 812 – 817 .

Möttönen T , Katisko J , Haapasalo J , et al. Defining the anterior nucleus of the thalamus (ANT) as a deep brain stimulation target in refractory epilepsy: delineation using 3 T MRI and intraoperative microelectrode recording . Neuroimage Clin . 2015 ; 7 : 823 – 829 .

Bonneville F , Welter ML , Elie C , et al. Parkinson disease, brain volumes, and subthalamic nucleus stimulation . Neurology . 2005 ; 64 ( 9 ): 1598 – 1604 .

Lee SH , Kim SS , Tae WS , et al. Regional volume analysis of the Parkinson disease brain in early disease stage: gray matter, white matter, striatum, and thalamus . AJNR Am J Neuroradiol . 2011 ; 32 ( 4 ): 682 – 687 .

O’Gorman RL , Jarosz JM , Samuel M , et al. CT/MR image fusion in the postoperative assessment of electrodes implanted for deep brain stimulation . Stereotact Funct Neurosurg . 2009 ; 87 ( 4 ): 205 – 210 .

Geevarghese R , O’Gorman Tuura R , Lumsden DE , et al. Registration accuracy of CT/MRI fusion for localisation of deep brain stimulation electrode position: an imaging study and systematic review . Stereotact Funct Neurosurg . 2016 ; 94 ( 3 ): 159 – 163 .

Holl EM , Petersen EA , Foltynie T , et al. Improving targeting in image-guided frame-based deep brain stimulation . Neurosurgery . 2010 ; 67 ( 2 Suppl Operative ): 437 – 447 .

Park SC , Lee JK , Kim SM , et al. Systematic stereotactic error reduction using a calibration technique in single-brain-pass and multitrack deep brain stimulations . Oper Neurosurg (Hagerstown) . 2018 ; 15 ( 1 ): 72 – 80 .

Kraff O , Quick HH . 7T: Physics, safety, and potential clinical applications . J Magn Reson Imaging . 2017 ; 46 ( 6 ): 1573 – 1589 .

Springer E , Dymerska B , Cardoso PL , et al. Comparison of routine brain imaging at 3 T and 7 T . Invest Radiol . 2016 ; 51 ( 8 ): 469 – 482 .

Forstmann BU , de Hollander G , van Maanen L , et al. Towards a mechanistic understanding of the human subcortex . Nat Rev Neurosci . 2016 ; 18 ( 1 ): 57 – 65 .

Abosch A , Yacoub E , Ugurbil K , Harel N . An assessment of current brain targets for deep brain stimulation surgery with susceptibility-weighted imaging at 7 tesla . Neurosurgery . 2010 ; 67 ( 6 ): 1745 – 1756 .

Dammann P , Kraff O , Wrede KH , et al. Evaluation of hardware-related geometrical distortion in structural MRI at 7 Tesla for image-guided applications in neurosurgery . Acad Radiol . 2011 ; 18 ( 7 ): 910 – 916 .

Hoff MN , McKinney A IV , Shellock FG , et al. Safety considerations of 7-T MRI in clinical practice . Radiology . 2019 ; 292 ( 3 ): 509 – 518 .

Yarach U , Luengviriya C , Stucht D , et al. Correction of B 0-induced geometric distortion variations in prospective motion correction for 7T MRI . MAGMA . 2016 ; 29 ( 3 ): 319 – 332 .

See AAQ , King NKK . Improving surgical outcome using diffusion tensor imaging techniques in deep brain stimulation . Front Surg . 2017 ; 4 : 54 .

Akram H , Sotiropoulos SN , Jbabdi S , et al. Subthalamic deep brain stimulation sweet spots and hyperdirect cortical connectivity in Parkinson’s disease . Neuroimage . 2017 ; 158 : 332 – 345 .

Horn A , Reich M , Vorwerk J , et al. Connectivity predicts deep brain stimulation outcome in Parkinson disease . Ann Neurol . 2017 ; 82 ( 1 ): 67 – 78 .

Coenen VA , Allert N , Paus S , et al. Modulation of the cerebello-thalamo-cortical network in thalamic deep brain stimulation for tremor: a diffusion tensor imaging study . Neurosurgery . 2014 ; 75 ( 6 ): 657 – 670 .

Coenen VA , Sajonz B , Reisert M , et al. Tractography-assisted deep brain stimulation of the superolateral branch of the medial forebrain bundle (slMFB DBS) in major depression . Neuroimage Clin . 2018 ; 20 : 580 – 593 .

Coenen VA , Allert N , Mädler B . A role of diffusion tensor imaging fiber tracking in deep brain stimulation surgery: DBS of the dentato-rubro-thalamic tract (drt) for the treatment of therapy-refractory tremor . Acta Neurochir (Wien) . 2011 ; 153 ( 8 ): 1579 – 1585 .

Sajonz BE , Amtage F , Reinacher PC , et al. Deep Brain Stimulation for Tremor Tractographic Versus Traditional (DISTINCT): study protocol of a randomized controlled feasibility trial . JMIR Res Protoc . 2016 ; 5 ( 4 ): e244 .

Chazen JL , Sarva H , Stieg PE , et al. Clinical improvement associated with targeted interruption of the cerebellothalamic tract following MR-guided focused ultrasound for essential tremor . J Neurosurg . 2018 ; 129 ( 2 ): 315 – 323 .

Plantinga BR , Temel Y , Duchin Y , et al. Individualized parcellation of the subthalamic nucleus in patients with Parkinson’s disease with 7T MRI . Neuroimage . 2018 ; 168 : 403 – 411 .

Johansen-Berg H , Behrens TE , Sillery E , et al. Functional-anatomical validation and individual variation of diffusion tractography-based segmentation of the human thalamus . Cereb Cortex . 2005 ; 15 ( 1 ): 31 – 39 .

Zrinzo L , Zrinzo LV , Tisch S , et al. Stereotactic localization of the human pedunculopontine nucleus: atlas-based coordinates and validation of a magnetic resonance imaging protocol for direct localization . Brain . 2008 ; 131 ( pt 6 ): 1588 – 1598 .

Elolf E , Bockermann V , Gringel T , et al. Improved visibility of the subthalamic nucleus on high-resolution stereotactic MR imaging by added susceptibility (T2*) contrast using multiple gradient echoes . AJNR Am J Neuroradiol . 2007 ; 28 ( 6 ): 1093 – 1094 .

Maruyama S , Fukunaga M , Fautz HP , et al. Comparison of 3T and 7T MRI for the visualization of globus pallidus sub-segments . Sci Rep . 2019 ; 9 ( 1 ): 18357 .

Li J , Li Y , Gutierrez L , et al. Imaging the centromedian thalamic nucleus using quantitative susceptibility mapping . Front Hum Neurosci . 2020 ; 13 : 447 .

Mai J , Majtanik M , Paxinos G . Atlas of the Human Brain . 4th ed. Elsevier Academic Press ; 2015 .

Edlow BL , Mareyam A , Horn A , et al. 7 Tesla MRI of the ex vivo human brain at 100 micron resolution . Sci Data . 2019 ; 6 ( 1 ): 244 .

Supplementary Materials

• Supplementary Table and Figures (PDF 2,851 KB)

Volume 135: Issue 5 (Nov 2021)

Illustration from Fan et al. (pp 1298–1309). Copyright Jun Fan. Published with permission.

• eTOC Alerts
• Ahead of Print Alerts
• Get Latest Issue TOC RSS

Article Information

Contributor notes.

* A.B. and A.L. contributed equally to this work.

INCLUDE WHEN CITING Published online March 26, 2021; DOI: 10.3171/2020.8.JNS201125.

Disclosures Dr. Zrinzo: consultant for Medtronic, Boston Scientific, and Elekta. Dr. Kalia: consultant for and honoraria from Medtronic. Dr. Fasano: grants, personal fees, and nonfinancial support from Abbvie, Medtronic, and Boston Scientific; personal fees from Sunovion, Chiesi Farmaceutici, and UCB; and grants and personal fees from Ipsen. Dr. Lozano: consultant for Medtronic, Boston Scientific, Abbott, and Insightec; and scientific director of Functional Neuromodulation.

• View raw image
• Download Powerpoint Slide

• PubMed Citation
• Articles by Alexandre Boutet
• Articles by Aaron Loh
• Articles by Clement T. Chow
• Articles by Alaa Taha
• Articles by Gavin J. B. Elias
• Articles by Clemens Neudorfer
• Articles by Jurgen Germann
• Articles by Michelle Paff
• Articles by Ludvic Zrinzo
• Articles by Alfonso Fasano
• Articles by Suneil K. Kalia
• Articles by Christopher J. Steele
• Articles by David Mikulis
• Articles by Walter Kucharczyk
• Articles by Andres M. Lozano
• Similar articles in PubMed

Google Scholar

• Article by Alexandre Boutet
• Article by Aaron Loh
• Article by Clement T. Chow
• Article by Alaa Taha
• Article by Gavin J. B. Elias
• Article by Clemens Neudorfer
• Article by Jurgen Germann
• Article by Michelle Paff
• Article by Ludvic Zrinzo
• Article by Alfonso Fasano
• Article by Suneil K. Kalia
• Article by Christopher J. Steele
• Article by David Mikulis
• Article by Walter Kucharczyk
• Article by Andres M. Lozano

© Copyright 1944-2024 American Association of Neurological Surgeons. All rights reserved, including those for text and data mining, AI training, and similar technologies.

Powered by PubFactory

• [185.66.14.133]
• 185.66.14.133

Character limit 500 /500

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List

An Update of the Possible Applications of Magnetic Resonance Imaging (MRI) in Dentistry: A Literature Review

Associated data.

Not applicable.

This narrative review aims to evaluate the current evidence for the application of magnetic resonance imaging (MRI), a radiation-free diagnostic exam, in some fields of dentistry. Background: Radiographic imaging plays a significant role in current first and second level dental diagnostics and treatment planning. However, the main disadvantage is the high exposure to ionizing radiation for patients. Methods: A search for articles on dental MRI was performed using the PubMed electronic database, and 37 studies were included. Only some articles about endodontics, conservative dentistry, implantology, and oral and craniofacial surgery that best represented the aim of this study were selected. Results: All the included articles showed that MRI can obtain well-defined images, which can be applied in operative dentistry. Conclusions: This review highlights the potential of MRI for diagnosis in dental clinical practice, without the risk of biological damage from continuous ionizing radiation exposure.

1. Introduction

The purpose of this narrative review was to evaluate the current evidence for the application of magnetic resonance imaging (MRI), a radiation-free diagnostic exam, in some fields of dentistry.

Radiographic imaging plays a significant role in the current first and second level of dental diagnostics and treatment planning [ 1 , 2 ]. With the introduction of cone beam computed tomography (CBCT), three-dimensional imaging prescription has become common in orthodontics, periodontology, implantology, and endodontics, with dedicated software becoming increasingly useful in these specific disciplines [ 3 ]. The merits of CBCT in treatment planning over conventional two-dimensional radiographic imaging are remarkable. However, the main disadvantage is the high exposure to ionizing radiation for patients, which does not allow clinicians to repetitively use this type of examination in a short window of time, with a consequent need for a careful assessment of the expected risk/benefit ratio in each individual case [ 4 ].

Three-dimensional images of the maxillofacial area are currently acquired by computed tomography (CT), cone beam computed tomography (CBCT), and magnetic resonance imaging (MRI) devices.

MRI, a well-established imaging technique in various areas of medicine, has become fundamental for the non-invasive diagnosis of soft tissue diseases since it has the great advantage of not using ionizing radiation, avoiding the biological damage related to the other three-dimensional imaging techniques such as CT and CBCT. MRI is almost comparable to the latter in terms of spatial resolution and data visualizing ability in the visions of the transverse and panoramic planes, which are most familiar to dentists [ 5 ].

For years, in dentistry, MRI imaging has always found a great application in the diagnosis of temporomandibular disorders (DTM) due to tissue histology, which has characteristics that perfectly match the type of diagnostic examination [ 6 ].

Therefore, considering the great amount of literature published regarding TMJ imaging, and since this diagnostic exam has been included in the guidelines for a long time, this review was focused on its applicability to the other branches of dentistry, which, to date, have not given great importance for this type of imaging technique [ 7 ]. Furthermore, since MRI does not use ionizing radiation, it is particularly relevant for repeated examinations in children [ 4 ].

Unlike radiographic imaging, the MRI technique is based on the presence of a magnetic field, formed by an MRI scanner in which the patient is positioned. The images are generated by measuring “signals” sent back by protons excited by the magnetic field, in particular by hydrogen atoms. Precisely for this reason, a better visualization of tissues containing water, such as the human brain, is obtained [ 8 ].

MRI creates the images using a strong and uniform static magnetic field and radio frequency pulses. When placed in a magnetic field, all substances are magnetized to a degree that depends on their magnetic susceptibility. Unfortunately, variations in magnetic field strength occurring at the interface between dental materials and adjacent tissues can lead to spatial distortions and signal loss, thus generating artifacts in the images [ 9 ].

In addition to the formation of artifacts, other undesirable effects of MRI could be radiofrequency, physical effects such as heating, and magnetically induced displacement (a mechanical effect) of dental materials [ 9 ].

Its application to dentistry, considering its risk–benefit ratio, would make it a very interesting exam. In particular, this review was focused on the comparison of the role of MRI in different branches of dentistry: endodontics, oral and maxillofacial surgery, and implantology. If RMI examination, to date, has not been considered enough as a diagnostic aid, the reason could be found in the disadvantages deriving from its application.

However, the high acquisition costs and long scan times are the main, and more discussed, disadvantages of the MRI technique. In addition, the risk that the patient suffers from claustrophobia, or that the patients have biomedical devices such as pacemakers, cochlear implants, neurostimulators, or infusion pumps represents an important contraindication; fixed metal prostheses and aneurism clips, to date, no longer represent a contraindication.

Chockattu et al. have clearly exposed the undesirable effects of magnetic resonance imaging [ 9 ]. The undesirable effects that are caused by the interaction of MRI and dental materials fall into three broad categories:

In addition to this typology, there are also artifacts caused by eddy currents, induced by alternating gradients and radiofrequency magnetic fields, which participate in generating distortions.

The complications related to RMI can cause malfunction, dislocation, and soft tissue burns (due to the absorption of radiofrequency energy).

Unwanted effects and the mechanisms that generate them are shown in Figure 1 .

Influence of an electromagnetic field.

Chockattu et al. emphasize the relevance of factors influencing unwanted effects, such as magnetic susceptibility and magnetic permeability. The susceptibility represents a measure of the extent to which a substance becomes magnetized when it is placed in an external magnetic field. The greater the magnetic permeability (and so, alloy composition) of a material, the more magnetic field distortion it will produce [ 13 ].

During MRI, the magnetic field could also be distorted by electric currents, due to electrical conductivity, flowing in materials within or close to the machine. These currents are induced in materials by a fluctuating magnetic field (summarized in Figure 1 ).

The magnetic permeability and tensile strength are linked by a relationship of direct proportionality. The tensile strength depends on the crystalline structure of the metal, so the past mechanical history of stainless-steel alloys determines their future effect on MRI images.

Moreover, the MRI sequence can also influence artifact formation; in effect, some sequences are more sensitive than others [ 14 ].

Dental materials, used in different dental procedures, which can often generate artifacts, do not include the different dental tissues, which do not generate artifacts.

There are many materials used in dentistry that can affect the quality of the MRI examination, which, in the past, represented a reason for not prescribing this diagnostic exam [ 10 , 12 ]. Endodontic materials such as resin-based sealer and gutta-percha seem not to produce detectable distortions on MRI [ 5 , 9 ]. In fixed orthodontic treatment, NiTi arch wires and stainless-steel brackets can distort local magnetic fields, causing large artifacts and making image interpretation very difficult [ 13 , 15 , 16 ]. Regarding maxillofacial prostheses, ferromagnetic devices should ideally be removed. Dental implants are made of titanium, a non-ferromagnetic material, and of ferromagnetic iron, which causes a drop-out of signal, causing artifacts [ 17 ].

In restorative dentistry, some materials seem to produce undetectable distortion on MR imaging, such as glass-ionomer cements and composite resins [ 5 ]. Polycarboxylate, zinc phosphate-based cement, and some modified dimethacrylates can also produce small image artifacts [ 18 ].

The amalgam, not frequently used today, that represented the gold standard in conservative dentistry until 20 years ago, is composed from several metals (copper, silver, tin, zinc, palladium, platinum, and mercury), with silver as the major component. Silver is a non-ferromagnetic metal, and so it does not have a significant influence in dental MRI.

In prosthetic dentistry, gold crowns are relatively free of ferromagnetic effects, due exclusively to traces of other metals that pollute the alloy. The ability to generate distorting ferromagnetic effects is very limited. Ceramic and zirconia crowns do not generate any artifacts, but, regarding zirconia, there are conflicting studies, with some comparing its effects to those of metals [ 19 , 20 ]. Metal–ceramic restorations, frequently with nickel alloys, seem to show a tendency to generate artifacts [ 9 , 15 ]. Dental materials that generate artifacts are summarized in Table 1 . It is also important to underline that not only the material, but also the shape of the metal object affects the quantity of artifacts: an arch- or ring-shaped implant will extinguish the signal inside the ring/arch, generating any artifacts.

Dental materials that generate artifacts.

2. Materials and Methods

A search for articles on dental MRI was performed using the PubMed electronic databases. The following keywords (Magnetic Resonance Imaging, MRI, Implantology, Endodontics, Periapical Lesions, Anatomy, Artifacts, Maxillary Sinus) combined with several Boolean operators were searched.

Five hundred and twenty-five articles were screened, and only 37 studies were included. According to the authors, only some articles about endodontics, conservative dentistry, implantology, and oral and craniofacial surgery that best represented the aim of this study were selected. The articles selected and not related to these branches of dentistry were considered only for the technical specifications and considerations on the functioning of MRI. Original research articles on gnathology and joint disorders, orthodontics, and prosthetics not concerning the topics listed above were excluded.

All the included articles showed that MRI can obtain well-defined images, which can be applied in operative dentistry. The studies’ selection flow-chart is represented in Figure 2 ; studies that passed the inclusion criteria and were considered for review are shown in Table 2 . In the right conditions, with proper attention to teeth, bone, and the tissues of the maxillofacial region, MRI can offer very important information not easily obtainable with other diagnostic exams. Dental MRI can also recognize pathological endodontic conditions such as decay, microcracks, and necrotic pulp tissues. Moreover, it can diagnose periapical granulomas from a cystic lesion and can represent an important aid in maxillary sinus conditions diagnosis and in implant surgery planning.

Selection process for the studies included (following PRISMA Statement) [ 21 ].

Studies included in the review.

4. Discussion

4.1. fundamental parameters in mri.

In dental MRI, signal-to-noise ratio (SNR) and resolution are two fundamental parameters to be considered; SNR is measured by calculating the ratio between the signal intensity in an area of interest and the standard deviation of the signal from the background [ 5 ]. The image resolution depends on the image voxel size. In MRI, the SNR can be improved by decreasing the matrix size, increasing the voxel size, increasing the field of view (FOV), reducing the bandwidth using surface coils, increasing the slice thickness, using an echo time (TE) of spin echo sequence as short as possible, and increasing the number of signal acquisitions (NA) [ 5 , 19 ].

The more the SNR is increased with the above-mentioned actions, the more the images’ definition decreases.

To reduce the FOV with the aim of increasing the resolution of the images without reducing SNR, it is necessary to use dedicated coils. The commonly used head or neck coils cannot reach an optimal resolution for being applied to improve dental diagnoses; intraoral positioning of the coil may increase both the resolution and the SNR, but it is very difficult to use due to anatomical limitations [ 9 , 19 ]. One of the most comfortable coil positions was proposed by Idiyatullin et al., with the advantages of using a loop coil, very similar to an impression tray, in the occlusal position for dental applications [ 22 ].

Despite the progress obtained so far, with the purpose of optimizing these parameters, the results (discussed below) are positive.

4.2. Apical Periodontitis Diagnosis

Regarding apical periodontitis, it is a chronic inflammatory disease of peri-radicular tissues, usually caused by a chronic bacterial infection of the root canal system near the bone. The pathogenesis of apical periodontitis and the cause of endodontic failure have been extensively reviewed by Siqueira and by Nair: the main role is played by bacteria (mainly obligate anaerobes and fungi), depending on the relationship with the host’s immune system [ 23 , 24 ]. Endo-osseous development of these conditions prevents the arrival of immunity cells and antibiotic molecules through the bloodstream. In order to be radiographically visible with bidimensional RX, a periapical radio-lucency should reach from 30% to 50% of bone mineral loss [ 25 ].

Sometimes these lesions heal spontaneously, sometimes they get worse, so much so that they enlarge and compress noble structures, or pour out as abscesses outside the bone [ 23 , 26 ].

The chronicity of these lesions makes them capable of corroding the bone in their proximity, visible radiographically as radiolucent lesions, although histological studies show that they can differ between granuloma or cyst [ 26 , 27 , 28 ].

Nair showed that up to 85% of all periapical lesions are granulomas [ 24 ].

Periapical granulomas contain granulomatous tissue, cell infiltrates, and a fibrous capsule, root cysts are considerably less frequent, and occur in two distinct histological categories: true apical cysts and pocket apical cysts [ 28 , 29 ]. True root cysts are entirely enclosed by the cyst wall epithelium, developed from the dormant epithelium, also known as epithelial rests of Malassez, after local inflammation stimuli. Periapical pocket cysts are lined by the epithelium but are open to the root canal, effectively isolating a pocket-like micro abscess from the periapical environment. This division is not merely histological, but has an important influence on the treatment, as the chances of recovery are very different from one to the other: granulomas and pocket cysts can heal after orthograde root canal therapy, while true cysts are self-sufficient and therefore less likely to be resolved with non-surgical treatment, hence, without removal of the cystic epithelium [ 26 , 28 ]. In the literature there are different opinions on this topic, not all authors agree about this definition [ 29 ]. Furthermore, larger lesions (more than 5 mm) are more likely to be root cysts associated with lower success rates for orthograde treatment [ 26 , 30 ].

From these considerations, the need for a diagnostic exam is highlighted, such as MRI, free from biological damage, unlike CBCT or CT, and able to evaluate in vivo the nature of the lesion and to orient the clinician towards the most appropriate treatment, whether it is surgical for true root cysts or endodontic, orthograde retreatment for periapical pocket cysts or granulomas [ 31 ].

One of the main advantages of MRI over CT and CBCT is the high soft tissue contrast and the ability to vary the contrast by changing the design of the MRI sequence, as well as the absence of ionizing radiation [ 30 ].

More specifically, MRI not only provides excellent soft tissue contrast but also allows for the evaluation of specific tissue components in different sequences.

Given these strengths, MRI has shown diagnostic superiority over CT techniques in various soft tissue associated pathologies in the head and neck region, in fact, MRI is the most suitable examination for the study of brain and solid tumors [ 9 ].

Therefore, surgical biopsy and subsequent histopathological evaluation remains the gold standard to confirm the diagnosis of different periapical lesions, but it obviously represents the most invasive technique considering the risk/benefit ratio. In this specific case, it is emphasized how this examination can simplify the diagnosis, having marked characteristics in evaluating lesions filled with liquid. Technical advances associated with the use of higher field strength, dedicated coil systems, and optimized sequencing techniques resulted in improved image quality, followed by increased interest in magnetic resonance imaging in dentistry [ 27 , 32 ].

To date, however, only Geibel and colleagues have systematically analyzed apical bone lesions with MRI; in a comparison between MRI and CBCT for the diagnosis of periapical lesions, they concluded that MRI is useful for the identification of fluids (hypointense T1-weighted images and hyperintense on T2-weighted images) and fibrous tissue (isointense on T1- and T2-weighted images) [ 32 ].

MRI has shown greater sensitivity in diagnosing periapical lesions than CBCT, in particular, when cystic fluid was present, thus excluding that it may be a vascularized lesion, such as a peri-apical granuloma. Moreover, it can more precisely diagnose the true dimensions of a lesion, and can provide a better estimation of the relationship between a lesion and critical structures, such as nerves and vessels [ 30 ].

Granulomas, on the other hand, are very heterogeneous due to the chronical infiltration of different immunity cells. Another important differentiation is represented by the wall of the lesion, with “thin-walled” cysts (mean: 1.6 mm) and “thick-walled” granulomas (mean: 4.6 mm), the latter also having poorly defined lesion margins in both MRI and in CBCT.

Moreover, the internal texture is very different; it is homogeneous in cysts, and inhomogeneous in granulomas [ 27 , 28 , 29 , 30 , 31 , 32 ]. Several authors postulated that dental MRI could detect inflammatory pathologies at an early stage, long before CBCT or conventional radiographs [ 14 , 30 , 32 ].

In many cases, the teeth on which these pathologies develop have already undergone primary endodontic treatment, and therefore the roots are reamed and filled with dense filling materials.

Geibel et al. believe it is very difficult to identify the root apex of the responsible tooth in these cases due to the presence of artifacts [ 32 ]. As previously stated, in the study of Chockattu et al., the same number of artifacts were not present with MRI, and when present, they appeared to produce undetectable distortions, unlike CBCT [ 9 ].

Therefore, it can be concluded that the MRI technique is essential for the analysis of periapical lesions, as these lesions must be adequately imaged with regard to resolution, contrast, signal-to-noise ratio, and susceptibility to artifacts.

4.3. Evaluation of Dental Fractures

Endodontically treated and incorrectly restored teeth, in addition to suffering more frequently from periapical infections, have a greater risk of fracturing [ 33 ].

Regarding dental fractures, MRI has the potential to help in determining the presence and extent of cracks and fractures in teeth due to good contrast, and especially without exposure to ionizing radiation as with CBCT, which is considered the current clinical standard [ 34 ].

In most cases, discontinuities cannot be definitively visualized in the absence of invasive measures such as CBCT imaging; in the study of Schuurmans et al., the aim was to develop MRI criteria for the identification of root cracks and fractures and to establish reliability and accuracy in their subsequent detection [ 5 , 34 ].

It is important to underline that these authors used in vivo MRI acquisition sequences on extracted teeth. A problem when evaluating these MRI studies is that in vitro sequences are frequently applied, with very long acquisition times, and they are impossible to be applied in vivo, and therefore are far from improving clinical practice. From the results of these studies, it is possible to highlight that MRI, thanks to the higher contrast, has allowed for better evaluation of cracks and fractures compared to CBCT imaging [ 34 ].

Part of these results is related to the reduced number of artifacts generated from radiopaque materials compared to CBCT imaging; this statement is very important because endodontically treated teeth that were root-filled are more prone to fracture if not correctly restored, due to tooth substance loss [ 9 , 10 , 33 ].

In conclusion, the advantages of contrast enhancement, and the absence or reduction of radiopaque material artifacts in MRI and comparable sensitivity and specificity measures with CBCT, suggest the importance of improvements in magnetic resonance quality, particularly in image acquisition and post-processing parameters. Always remembering the absence of ionizing radiation, and the continuous improvements that this imaging exam is obtaining, the next applications of dental MRI in detecting dental cracks or fractures may involve defining the minimum physical size for detection using advanced MRI sequences [ 12 , 18 , 19 ].

4.4. Endodontics, Endodontic Anatomy and Conservative Dentistry

Regarding endodontic anatomy, while performing an endodontic treatment, it is extremely important to create a correct and accurate topographic image of the root canal system; knowing the anatomy well before starting endodontic treatment allows the clinician to use the most suitable instruments in the correct way, avoiding subjecting them to considerable stress that could lead to intracanal separation [ 35 , 36 , 37 , 38 ].

Up to date, visualization of root canal topography and dental anatomy has been obtained by conventional bi-dimensional radiographs, and only in recent years has CBCT been increasingly applied, due to the reduction of the exposure dose, the increasing availability of the machinery in the private practice, and the reduced costs compared to the past or to other exams. MRI offers high-level tissue visibility, equal to or even greater than CT and CBCT, but it requires sufficient resolution that tends to be achieved only with much longer scan times, without, however, exposure to ionizing radiation.

Several articles have shown the usefulness of spin echo and gradient echo imaging, single point imaging, and SPRITE and STRAFI techniques for the visualization of tooth surface geometry, as well as for distinguishing between soft tissue and mineralized tissue in extracted teeth [ 38 ]. The high-intensity signal from water and the lack of signal from mineralized tissues produce a high contrast that allows for the recognition of the dental crown and the outline of the pulp chamber, root canals, and carious lesion [ 39 , 40 , 41 ].

Bracher et al. stated that carious tissues provide an intense signal, easily recognizable in the 3D reconstruction performed by the software.

In order for magnetic resonance imaging to be applied to endodontic clinical practice, it is necessary to scan at the microscopic level, with microscopy MRI defined as an MRI with voxel resolutions better than 100 mm 3 . Magnetic resonance microscopy chambers are generally small, typically less than 1 cm3. With a resolution of about 100–300 mm, magnetic resonance microscopy could lead to a better understanding of processes that occur inside the teeth.

The obtained microscopic images allow for adequate visualization of the pulp chamber, pulp, and root canals. Ploder et al. used a magnetic resonance exam as an imaging examination complementary to the electrical pulp test for the evaluation of pulp health and of pathological processes occurring within the dental pulp tissue [ 42 ].

After pulp health determination with the electrical test, healthy pulp could show a signal on T2-weighted images ranged between intermediate and high hyperintense values, which becomes shorter according to patient age, due to secondary dentin accumulation [ 41 , 42 ].

The characteristic of magnetic resonance represents tissues that are rich in water very well, for this reason, the inflammatory response, which develops edema, will be evaluated in an ideal way, and certainly better than dental necrosis, in which we expect a loss in the content of water in the pulp [ 42 ].

MRI can therefore be useful in evaluating reperfusion, for example, that concerning regenerative endodontic procedures (REPs) and dental trauma [ 42 , 43 , 44 , 45 ].

The application limit of this examination is that, to obtain a sufficient resolution for clinical evaluation in vivo, it takes up to 90 min. It is expected that, with technological development, the imaging time will be reduced in the future, making it fast enough to facilitate clinical use.

The visualization of hard tissues, such as enamel and dentin that do not have MRI signals, considering the low content of protons, remains the real technical challenge to be faced in making MRI a daily diagnostic exam in dentistry [ 46 ].

The presented results show the feasibility of using magnetic resonance microscopy to carefully visualize root canal anatomy, applicable for the planning of endodontic procedures while avoiding NiTi rotary instruments, intracanal separation, or other iatrogenic errors, without having an increase in radiation-related biological risks [ 39 , 47 ].

4.5. Implantology

Regarding implantology, the aim of the study proposed by Probst et al. was to show whether computer-aided 3D implant planning with template-guided positioning of dental implants based on MRI data is a clinically valid procedure [ 48 ].

It is very important to point out that all cases in this study were performed by guided implant surgery, virtually planned based on MRI and intraorally transferred by static guides. It is necessary to underline that the authors have reported a deviation between the virtually planned implant position and the resulting final implant position, a deviation of occlusal surfaces between the digitized and occlusal plaster models derived from the MRI data, and the visualization of important anatomical structures that was completely acceptable for clinical application. It is, therefore, possible to define that MR images are sufficiently accurate to show all anatomical structures relevant to dental implant planning, free from ionizing radiation, with an excellent risk/benefit ratio.

In the typical MRI representation, hard dental tissues and bone tissue appear extremely dark due to the poor liquid composition. However, by inverting the dark signal values of the MR image datasets, it is possible to provide a bright or white color to the teeth and various bone structures, and so, an image more similar to CBCT is obtained [ 12 , 13 , 48 ].

The sequence parameters have been optimized considering the spatial resolution and total image acquisition time requirements; therefore, the longer the image acquisition time, the greater the chance of motion artifacts occurring.

The aforementioned authors suggest that the isotropic 3D size with a 0.6 mm3 voxel resulted in a reasonable acquisition time of just over 3:08 min.

In implantology it is very important to consider the anatomical limitations. For example, the mandibular canal position, an extremely important limitation in the posterior atrophic mandible, is excellently displayed with the use of the T1-weighted 3D sequence.

In the absence of the cortical bone lining the mandibular canal, or in the presence of metal restorations near the inferior alveolar nerve, artifacts can make its location very difficult. This unfortunate event occurs in both T1-weighted sequences and in CBCT imaging [ 9 , 48 ].

However, MR imaging offers a unique advantage and added value through the application of soft tissue contrast in specific sequences. While the T1-weighted sequence is practically a “bone sequence”, and therefore comparable to CBCT imaging, the T2-weighted STIR sequence can work as a “soft tissue and nerve sequence” during implant planning, which allows for direct nerve and blood vessel imaging [ 19 , 20 , 48 , 49 , 50 ].

With increasingly adequate programming software, it will be possible to obtain more information from both sequences in order to improve implant programming, always with an excellent risk/benefit ratio, considering the absence of exposure to ionizing radiation.

One of the most complex problems to be solved is represented by motion artifacts, which can compromise the overall image quality of MR imaging due to the significantly increased examination times compared to CT or CBCT, which represents the major limitation nowadays [ 5 , 7 , 9 , 48 ].

This problem could be solved by trying to reduce the examination time, increasing the stability of the patient’s head, and using more effective software to digitally correct these artifacts. However, there is always the problem of artifacts due to the presence of metallic materials, which can affect the representation of important structures when in proximity [ 9 , 10 , 11 , 12 , 13 , 14 ].

Except for titanium plates and synthesis screws, artifacts due to the presence of metal dental restorations were limited to the occlusal plane area, and therefore minimally limit the implant treatment plan.

The presence of artifacts of the occlusal plane can represent a limit just when a tooth-supported template-guide is produced only from the MRI exam. Other anatomical structures such as bones, the maxillary sinus, and soft tissues were substantially unaffected, not compromising implant planning at all [ 48 ].

Artifacts also represent an important problem for CBCT and CT examinations, considering, moreover, the biological damage that these examinations generate.

However, while some materials such as stainless steel and cobalt–chromium alloy are responsible for pronounced artifacts, both in CBCT and in MRI, that may no longer allow for a reasonable diagnosis, the majority of dental materials such as zirconia, amalgams, gold alloys, gold–ceramic crowns, titanium alloys, some composites, and nickel–titanium cause artifacts in a minor way [ 9 , 16 , 17 , 18 , 19 , 20 ].

An interesting evaluation regarding the article of Probst et al. is that the tooth-supported templates were obtained exclusively with images from MRI, and not from intraoral scanners or other types of imaging or impressions, thus representing a valid alternative, with excellent clinical results.

The 3D comparisons of deviations between MRI reconstructed and scan-derived tooth surfaces, carried out for further evaluation of the methodology, showed acceptable values for clinical application [ 20 , 48 ].

The study emphasizes that these results were achieved with a maximum number of 5–6 metal restorations per jaw [ 48 ].

It must be considered, however, that a tooth that has undergone artifacts can be excluded from the template, placing it on all available nearby ones.

In addition, MRI can provide added diagnostic value due to the excellent soft tissue contrast, which allows, for example, a direct image of peripheral nerve tissue, such as the inferior alveolar nerve, useful for implant planning, as demonstrated in this study [ 24 , 49 , 50 ].

In radiographic imaging, the problem of artefacts is always present, but peri-implant bone defect evaluation, or studies about bone morphology near the implant, are still being carried out [ 51 ].

Moreover, radiographic imaging is used for patient follow-up, but always with exposure to a certain dose of ionizing radiation. Precisely from this perspective, magnetic resonance imaging can become an easily repeatable diagnostic test with an excellent risk/benefit ratio.

In the context of implant surgery, magnetic resonance imaging allows for the detailed measurement of mucosal thickness and can aid in the planning of palatal tissue harvesting to obtain soft tissue augmentation [ 52 ].

Despite the various disadvantages that characterize this method, the possibility of being able to perform an examination with a very low risk/benefit ratio is of truly unparalleled value, which must lead to a greater interest in the development of this diagnostic exam.

4.6. Maxillary Sinus Diagnosis and Surgery

Regarding the maxillary sinus, the evaluations made by Aktuna Belgin et al. and Dong et al. underlined its importance, showing how it is a structure that can be well studied by MRI, as also pointed out by Panou et al. and Özdemir et al. [ 53 , 54 , 55 , 56 , 57 ].

Successful treatment of sino-nasal disorders, complete knowledge, and correct visualization of the anatomical conditions of the osteomeatal complex and paranasal sinuses is fundamental in head, neck, and oral-maxillofacial surgery. The maxillary sinuses are very interesting in dental clinical practice, and very frequently studied for atrophic jaw rehabilitation [ 53 ]. For this reason, it is possible to affirm that they represent both a limitation and a frequent rehabilitation possibility. Knowledge of the anatomical variables of the maxillary sinus is precious to prevent possible accidents and complications in maxillofacial surgery, as well as in the preoperative evaluation in dental implant treatment or in more complex bone regenerations.

Previous studies have also examined volumetric changes in the maxillary sinus; relationships with tooth position; and orthodontic treatment-induced changes such as rapid expansion, septal deviation, and sinus pathologies, as well as examining the differences in the size and anatomy of the maxillary sinus based on age, sex, and race [ 53 , 56 , 57 ].

Published studies on maxillary sinus volume have produced differing results. Rani et al. found no significant differences in volume (MSV) between the left and right maxillary sinus, and reported that MSV was significantly higher in males than in females [ 57 ].

This finding corroborates the findings of the study of Aktuna Belgin et al.; Özdemir et al. and Butaric et al. stated that maxillary sinus development continues into the second and third decade of life in females and males, respectively, with an age-associated decrease in volume occurring after development is completed. All these results have been confirmed by Rani et al. [ 55 , 56 , 57 , 58 ].

Compared to CT and CBCT, MRI has fewer metallic artifacts, but with longer exam execution times, and can be used with 3D medical imaging software, allowing for the examination of images obtained in the axial, coronal, and sagittal planes, showing images very similar to that of CBCT, also by the use of specific filters, to increase the contrast between different structures [ 9 , 11 , 12 , 13 ].

In maxillary sinus surgery for implant placement, it is important to know and visualize the state of the Schneiderian membrane and any reactive thickening phenomena [ 59 , 60 ]. CBCT allows for visualizing the three-dimensional bone morphology, but the mucosa is poorly defined, despite exposure to ionizing radiation [ 2 , 4 , 5 ].

In this evaluation, MRI is positioned as a very interesting exam with great margins for improvement, having also demonstrated its usefulness in complete implant planning and in defining the state of health of the maxillary sinus and the Schneiderian membrane for any bone regeneration [ 60 , 61 ].

Moreover, it is necessary to consider that many imaging systems are undergoing considerable changes due to the continuous development of methods that exploit artificial intelligence (AI).

The development of artificial intelligence (AI) technology has proven to be successful in many research fields of medical imaging and various applications of robotic surgery. It can be extremely useful in recognizing landmarks in MRI and optimizing the image produced. In the near future, as with many kinds of software, we expect applications of these technologies to magnetic resonance too, in order to improve and make the use of this interesting diagnostic exam, free from ionizing radiation, more suitable for clinical dental practice [ 62 , 63 , 64 ].

5. Conclusions

With the development of technology, as happened to CT and CBCT in the past, software programs that perform three-dimensional modelling have been introduced in MRI.

As pointed out, the three-dimensional modelling can be excellently applied in the measurement of maxillary sinus volume and Schneiderian membrane thickness, to decide which rehabilitation is the most suitable.

Moreover, MRI-based computer-assisted implant surgery is demonstrated to be a feasible and accurate procedure, eliminating radiation exposure.

In addition, MRI, compared to the CBCT, better allows for the diagnostic visualization of soft tissues such as the alveolar inferior nerve, which is the most important limitation in the context of posterior mandible rehabilitation.

MRI could become a more common diagnostic exam, both in research and clinical endodontics, providing the possibility to evaluate decay extensions, vitality, and vascularization of the pulp after trauma or after regenerative endodontics; the presence of soft tissue remnants after endodontic procedures or the early detection of missing canals, cracks, and fractures; and precise follow-up of periapical lesions, with the great advantage of avoiding the risk of ionizing radiation damage.

The development of this method can really allow for an improvement in the diagnosis and prognosis of periapical bone lesions.

The main disadvantages of this examination remain the difficult visualization of tissues poor in water, which, however, has proven to be correctable by dedicated software, and can lead to excellent results. Patients suffering from claustrophobia, the presence of devices that prevent the examination from taking place, artifacts from materials and movements, the cost, the lack of availability, and the long examination time represent the disadvantages that will need to be improved in the future.

The results analyzed in this review highlight the potential of MRI for diagnosis in dental clinical practice, without the risk of biological damage from continuous ionizing radiation exposure.

Author Contributions

Conceptualization, R.R. and A.Z.; methodology, A.C.; validation, D.D.N.; formal analysis, L.T.; investigation, R.R.; resources, A.M.; data curation, A.M.; writing—original draft preparation, R.R.; writing—review and editing, R.R. and A.Z.; visualization, L.T.; supervision, L.T.; All authors have read and agreed to the published version of the manuscript.

This research received no external funding.

Institutional Review Board Statement

Informed consent statement, data availability statement, conflicts of interest.

The authors declare no conflict of interest.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

• Open access
• Published: 24 July 2023

Deep learning applications to breast cancer detection by magnetic resonance imaging: a literature review

• Richard Adam 1 ,
• Kevin Dell’Aquila 1 ,
• Laura Hodges 1 ,
• Takouhie Maldjian 1 &
• Tim Q. Duong 1  

Breast Cancer Research volume  25 , Article number:  87 ( 2023 ) Cite this article

11k Accesses

15 Citations

1 Altmetric

Metrics details

Deep learning analysis of radiological images has the potential to improve diagnostic accuracy of breast cancer, ultimately leading to better patient outcomes. This paper systematically reviewed the current literature on deep learning detection of breast cancer based on magnetic resonance imaging (MRI). The literature search was performed from 2015 to Dec 31, 2022, using Pubmed. Other database included Semantic Scholar, ACM Digital Library, Google search, Google Scholar, and pre-print depositories (such as Research Square). Articles that were not deep learning (such as texture analysis) were excluded. PRISMA guidelines for reporting were used. We analyzed different deep learning algorithms, methods of analysis, experimental design, MRI image types, types of ground truths, sample sizes, numbers of benign and malignant lesions, and performance in the literature. We discussed lessons learned, challenges to broad deployment in clinical practice and suggested future research directions.

Breast cancer is the most common cancer and the second leading cause of cancer death in women. One in eight American women (13%) will be diagnosed with breast cancer in their lifetime, and one in 39 women (3%) will die from breast cancer (American Cancer Society Statistics, 2023). The American Cancer Society recommends yearly screening mammography for early detection of breast cancer for women, which may begin at age 40 [ 1 ]. About 2%–5% of women in the general population in the US have a lifetime risk of breast cancer of 20% or higher [ 1 ], although it can vary depending on the population being studied and the risk assessment method used. The ACS recommends yearly breast magnetic resonance imaging (MRI) in addition to mammography for women with 20–25% or greater lifetime risk [ 1 ]. Early detection and treatment are likely to result in better patient outcomes.

MRI is generally more sensitive and offers more detailed pathophysiological information but is less cost effective compared to mammography for population-based screening [ 2 , 3 ]. Breast MRI utilizes high-powered magnets and radio waves to generate 3D images. Cancer yield from MRI alone averages 22 cancers for every 1000 women screened, a rate of cancer detection roughly 10 times that achieved with screening mammography in average-risk women, and roughly twice the yield achieved with screening mammography in high-risk women [ 4 ]. Many recent studies have established contrast-enhanced breast MRI as a screening modality for women with a hereditary or familial increased risk for the development of breast cancer [ 5 ].

Interpretation of breast cancer on MRI relies on the expertise of radiologists. The growing demand for breast MRI and the shortage of radiologists has resulted in increased workload for radiologists [ 6 , 7 ], leading to long wait times and delays in diagnosis [ 8 , 9 ]. Machine learning methods show promise in assisting radiologists, in improving accuracy with the interpretation of breast MRI images and supporting clinical decision-making and improving patient outcomes [ 10 , 11 ]. By analyzing large datasets of MRIs, machine learning algorithms can learn to identify and classify suspicious areas, potentially reducing the number of false positives and false negatives [ 11 , 12 ] and thus improving diagnostic accuracy. A few studies have shown that machine learning can outperform radiologists in detecting breast cancer on MRIs [ 13 ]. Machine learning could also help to prioritize worklists in a radiology department.

In recent years, deep learning (DL) methods have revolutionized the field of computer vision with wide range of applications, from image classification and object detection to semantic segmentation and medical image analysis [ 14 ]. Deep learning is superior to traditional machine learning because of its ability to learn from unstructured or unlabeled data [ 14 ]. Unlike traditional machine algorithms which require time-consuming data labeling, deep learning algorithms are more flexible and adaptable as they can learn from data that are not labeled or structured [ 15 ]. There have been a few reviews on deep learning breast cancer detection. Oza et al. reviewed detection and classification on mammography [ 16 ]. Saba et al. [ 17 ] presented a compendium of state-of-the-art techniques for diagnosing breast cancers and other cancers. Hu et al. [ 18 ] provided a broad overview on the research and development of artificial intelligence systems for clinical breast cancer image analysis, discussing the clinical role of artificial intelligence in risk assessment, detection, diagnosis, prognosis, and treatment response assessment. Mahoro et al. [ 10 ] reviewed the applications of deep learning to breast cancer diagnosis across multiple imaging modalities. Sechopoulos et al. [ 19 ] discussed the advances of AI in the realm of mammography and digital tomosynthesis. AI-based workflows integrating multiple datastreams, including breast imaging, can support clinical decision-making and help facilitate personalized medicine [ 20 ]. To our knowledge, there is currently no review that systematically compares different deep learning studies of breast cancer detection using MRI. Such a review would be important because it could help to delineate the path forward.

Figure  1 shows a graphic representation of a deep learning workflow. The input layer represents the breast cancer image that serves as input to the CNN. The multiple convolutional layers are stacked on top of the input layer. Each convolutional layer applies filters or kernels to extract specific features from the input image. These filters learn to detect patterns such as edges, textures, or other relevant features related to breast cancer. After each convolutional layer, activation functions like rectified linear unit (ReLU) are typically applied to introduce nonlinearity into the network. Following some of the convolutional layers, pooling layers are used to downsample the spatial dimensions of the feature maps. Common pooling techniques include max-pooling or average pooling. Pooling helps reduce the computational complexity and extract the most salient features. After the convolutional and pooling layers, fully connected layers are employed. These layers connect all the neurons from the previous layers to the subsequent layers. Fully connected layers enable the network to learn complex relationships between features. The final layer is the output layer, which provides the classification or prediction. In the case of breast cancer detection, it might output the probability or prediction of malignancy or benignity.

The input layer represents the breast cancer image that serves as input to the CNN. The multiple convolutional layers are stacked on top of the input layer. Pooling layers are used to downsample the spatial dimensions of the feature maps. Fully connected layers are then employed to connect all the neurons from the previous layers to the subsequent layers. The final layer is the output layer, which provides the classification

The goal of this study was to review the current literature on deep learning detection of breast cancer using breast MRI. We included literature in which DL was used for both primary screening setting and as a supplemental detection tool. We compared different deep learning algorithms, methods of analysis, types of ground truths, sample size, numbers of benign and malignant lesions, MRI image types, and performance indices, among others. We also discussed lessons learned, challenges of deployment in clinical practice and suggested future research directions.

Materials and methods

No ethics committee approval was required for this review.

Search strategy and eligibility criteria

PRISMA guidelines for reporting were adopted in our systematic review. The literature search was performed from 2017 to Dec 31, 2022, using the following key words: “breast MRI,” “breast magnetic resonance imaging,” “deep learning,” “breast cancer detection,” and “breast cancer screening.” The database included Pubmed, Semantic Scholar, ACM Digital Library, Google search, Google Scholar, and pre-print depositories (such as Research Square). We noted that many of the computing or machine learning journals were found on sites other than Pubmed. Some were full-length peer-reviewed conference papers, in contrast with small conference abstracts. Articles that were not deep learning (such as texture analysis) were excluded. Only original articles written in English were selected. Figure  2 shows the flowchart demonstrating how articles were included and excluded for our review. The search and initial screening for eligibility were performed by RA and independently verified by KD and/or TD. This study did not review DL prediction of neoadjuvant chemotherapy which has recently been reviewed [ 21 ].

PRISMA selection flowchart

Pubmed search yielded 59 articles, of which 22 were review articles, 30 were not related to breast cancer detection on MRI, and two had unclear/unconventional methodologies. Five articles were found in Pubmed search after exclusion (Fig.  2 ). In addition, 13 articles were found on different databases outside of Pubmed, because many computing and machine learning journals were not indexed by Pubmed. A total of 18 articles were included in our study (Table 1 ). Two of the studies stated that the patient populations were moderate/high risk [ 22 , 23 ] or high risk [ 23 ], while the remaining papers do not state whether the dataset was from screening or supplemental MRI.

In this review, we first summarized individual papers and followed by generalization of lessons learned. We then discussed challenges of deployment in the clinics and suggested future research directions.

Summary of individual papers

Adachi et al. [ 13 ] performed a retrospective study using RetinaNet as a CNN architecture to analyze and detect breast cancer in MIPs of DCE fat-suppressed MRI images. Images of breast lesions were annotated with a rectangular region-of-interest (ROI) and labeled as “benign” or “malignant” by an experienced breast radiologist. The AUCs, sensitivities, and specificities of four readers were also evaluated as well as those of readers combined with CNN. RetinaNet alone had a higher area under the curve (AUC) and sensitivity (0.925 and 0.926, respectively) than any of the readers. In two cases, the AI system misdiagnosed normal breast as malignancy, which may be the result of variations in normal breast tissue. Invasive ductal carcinoma near the axilla was missed by AI, possibly due to confusion for normal axillary lymph node. Wider variety of data and larger datasets for training could alleviate these problems.

Antropova et al. [ 24 ] compared MIP derived from the second post-contrast subtraction T1-weighted image to the central slice of the second post-contrast image with and without subtraction. The ground truth was ROIs based on radiology assessment with biopsy-proven malignancy. MIP images showed the highest AUC. Feature extraction and classifier training for each slice for DCE-MRI sequences, with slices in the hundreds, would have been computationally expensive at the time. MIP images, in widespread use clinically, contain enhancement information throughout the tumor volume. MIP images, which represent a volume data, avoid using a plethora of slices, and are, therefore, faster and computationally less intensive and less expensive. MIP (AUC = 0.88) outperformed one-slice DCE image, and subtracted DCE image (AUC = 0.83) outperformed single-slice DCE image (AUC = 0.80). The subtracted DCE image is derived from 2 timepoints, the pre-contrast image subtracted from the post-contrast image, which produces a higher AUC. Using multiple slices and/or multiple timepoints could further increase the AUC with DCE images, possibly even exceeding that of the MIP image (0.88). This would be an area for further exploration.

Ayatollahi et al. [ 22 ] performed a retrospective study using 3D RetinaNet as a CNN architecture to analyze and detect breast cancer in ultrafast TWIST DCE-MRI images. They used 572 images (365 malignant and 207 benign) taken from 462 patients. Bounding boxes drawn around the lesion in the images were used as ground truth. They found a detection rate of 0.90 and a sensitivity of 0.95 with tenfold cross validation.

Feng et al. [ 23 ] implemented the Knowledge-Driven Feature Learning and Integration model (KFLI) using DWI and DCE-MRI data from 100 high-risk female patients with 32 benign and 68 malignant lesions, segmented by two experienced radiologists. They reported 0.85 accuracy. The model formulated a sequence division module and adaptive weighting module. The sequence division module based on lesion characteristics is proposed for feature learning, and the adaptive weighting module proposed is used for automatic feature integration while improving the performance of cooperative diagnosis. This model provides the contribution of sub-sequences and guides the radiologists to focus on characteristic-related sequences with high contribution to lesion diagnosis. This can save time for the radiologists and helps them to better understand the output results of the deep networks. As such, it can extract sufficient and effective features from each sub-sequence for a comprehensive diagnosis of breast cancer. This model is a deep network and domain knowledge ensemble that achieved high sensitivity, specificity, and accuracy.

Fujioka et al. [ 25 ] used 3D MIP projection from early phase (1–2 min) of dynamic contrast-enhanced axial fat-suppressed DCE mages, with performance of CNN models compared to two human readers (Reader 1 = breast surgeon with 5 years of experience and Reader 2 = radiologist with 20 years of experience) in distinguishing benign from malignant lesions. The highest AUC achieved with deep learning was with InceptionResNetV2 CNN model, at 0.895. Mean AUC across the different CNN models was 0.830, and range was 0.750–0.895, performing comparably to human readers. False-positive masses tended to be relatively large with fast pattern of strong enhancement, and false-negative masses tended to be relatively small with medium to slow pattern of enhancement. One false positive and one false negative for non-mass enhancing lesion that was observed were also incorrectly diagnosed by the human readers. The main limitation of their study was small sample size.

Haarburger et al. [ 26 ] performed an analysis of 3D whole-volume images on a larger cohort ( N  = 408 patients), yielding an AUC of up to 0.89 and accuracy of 0.81, further establishing the feasibility of using 3D DCE whole images. Their method involved feeding DCE images from 5 timepoints (before contrast and 4 times post-contrast) and T2-weighted images to the algorithms. The multicurriculum ensemble consisted of a 3D CNN that generates feature maps and a classification component that performs classification based on the aggregated feature maps made by the previous components. AUC range of 0.50–0.89 was produced depending on the CNN models used. Multiscale curriculum training improved simple 3D ResNet18 from an AUC of 0.50 to an AUC of 0.89 (ResNet18 curriculum). A radiologist with 2 years of experience demonstrated AUC of 0.93 and accuracy of 0.93. An advantage of the multicurriculum ensemble is the elimination of the need for pixelwise segmentation for individual lesions, as only coarse localization coordinates for Stage 1 training (performed in 3D in this case) and one global label per breast for Stage 2 training is needed, where Stage 2 involved predictions of whole images in 3D in this study. The high performance of this model can be attributed to the high amount of context and global information provided. Their 3D data use whole breast volumes without time-consuming and cost prohibitive lesion segmentation. A major drawback of 3D images is the requirement of more RAM and many patients required to train the model.

Herent et al. [ 27 ] used T1-weighted fat-suppressed post-contrast MRI in a CNN model that detected and then characterized lesions ( N  = 335). Lesion characterization consisted of diagnosing malignancy and lesion classification. Their model, therefore, performed three tasks and thereby was a multitask technique, which limits overfitting. ResNET50 neural network performed feature extraction from images, and images were processed by the algorithm’s attention block which learned to detect abnormalities. Images were fed into a second branch where features were averaged over the selected regions, then fitted to a logistic regression to produce the output. On an independent test set of 168 images, a weighted mean AUC of 0.816 was achieved. The training dataset consisted of 17 different histopathologies, of which most were represented as very small percentages of the whole dataset of 335. Several of the listed lesion types represented less than 1% of the training dataset. This leads to the problem of overfitting. The authors mention that validation of the algorithm by applying it to 3D images in an independent dataset, rather than using the single 2D images as they did, would show if the model is generalizable. The authors state that training on larger databases and with multiparametric MRI would likely increase accuracy. This study shows good performance of a supervised attention model with deep learning for breast MRI.

Hu et al. [ 28 ] used multiparametric MR images (DCE-MRI sequence and a T2-weighted MRI sequence) in a CNN model including 616 patients with 927 unique breast lesions, 728 of which were malignant. A pre-trained CNN extracted features from both DCE and T2w sequences depicting lesions that were classified as benign or malignant by support vector machine classifiers. Sequences were integrated at different levels using image fusion, feature fusion, and classifier fusion. Feature fusion from multiparametric sequences outperformed DCE-MRI alone. The feature fusion model had an AUC of 0.87, sensitivity of 0.78, and specificity of 0.79. CNN models that used separate T2w and DCE images into combined RBG images or aggregates of the probability of malignancy output from DCE and T2w classifiers both did not perform significantly better than the CNN model using DCE-alone. Although other studies have demonstrated that single-sequence MRI is sufficient for high CNN performance, this study demonstrates that multiparametric MRI (as fusion of features from DCE-MRI and T2-weighted MRI) offers enough information to outperform single-sequence MRI.

Li et al. [ 29 ] used 3D CNNs in DCE-MR images to differentiate between benign and malignant tumors from 143 patients. In 2D and 3D DCE-MRI, a region-of-interest (ROI) and volume-of-interest (VOI) were segmented, and enhancement ratios for each MR series were calculated. The AUC value of 0.801 for the 3D CNN was higher than the value of 0.739 for 2D CNN. Furthermore, the 3D CNN achieved higher accuracy, sensitivity, and specificity values of 0.781, 0.744, and 0.823, respectively. The DCE-MRI enhancement maps had higher accuracy by using more information to diagnose breast cancer. The high values demonstrate that 3D CNN in breast cancer MR imaging can be used for the detection of breast cancer and reduce manual feature extraction.

Liu et al. [ 30 ] used CNN to analyze and detect breast cancer on T1 DCE-MRI images from 438 patients, 131 from I-SPY clinical trials and 307 from Columbia University. Segmentation was performed through an automated process involving fuzzy C-method after seed points were manually indicated. This study included analysis of commonly excluded image features such as background parenchymal enhancement, slice images of breast MRI, and axilla/axillary lymph node involvement. The methods also minimized annotations done at pixel level, to maximize automation of visual interpretation. These objectives increased efficiency, decreased subjective bias, and allowed for complete evaluation of the whole image. Obtaining images with multiple timepoints from multiple institutions made the algorithm more generalizable. The CNN model achieved AUC of 0.92, accuracy of 0.94, sensitivity of 0.74, and specificity of 0.95.

Marrone et al. [ 31 ] used CNN to evaluate 42 malignant and 25 benign lesions in 42 women. ROIs were obtained by an experienced radiologist, and manual segmentation was performed. Accuracy of up to 0.76 was achieved. AUC as high as 0.76 was seen on pre-trained AlexNet versus 0.73 on fine-tuning of pre-trained AlexNet where the last trained layers were replaced by untrained layers. The latter method could allow reduced number of training images needed. The training from scratch AlexNet model is accomplished when AlexNet pre-trained on the ImageNet database is used to extract a feature vector from the last internal CNN layer, and a new supervised training is employed, which yielded the lowest AUC of 0.68 and accuracy of 0.55.

Rasti et al. [ 32 ] analyzed DCE-MRI subtraction images from MRI studies ( N  = 112) using a multi-ensemble CNN (ME-CNN) functioning as a CAD system to distinguish benign from malignant masses, producing 0.96 accuracy with their method. The ME-CNN is a modular and image-based ensemble, which can stochastically partition the high-dimensional image space through simultaneous and competitive learning of its modules. It also has the advantages of fast execution time in both training and testing and a compact structure with a small number of free parameters. Among several promising directions, one could extend the ME-CNN approach to the pre-processing stage, by combining ME-CNN with recent advances in fully autonomous CNNs for semantic segmentation.

Truhn et al. [ 33 ] used T2-weighted images with one pre-contrast and four post-contrast DCE images in 447 patients with 1294 enhancing lesions (787 malignant and 507 benign) manually segmented by a breast radiologist. Deep learning with CNN demonstrated an AUC of 0.88 which was inferior to prospective interpretation by one of the three breast radiologists (7–25 years of experience) reading cases in equal proportion (0.98). When only half of the dataset was used for training ( n  = 647), the AUC was 0.83. The authors speculate that with increased training on a greater number of cases that their model could improve its performance.

Wu et al. [ 34 ] trained a CNN model to analyze and detect lesions from DCE T1-weighted images from 130 patients, 71 of which had malignant lesions and 59 had benign tumors. Fuzzy C-means clustering-based algorithm automatically segmented 3D tumor volumes from DCE images after rectangular region-of-interest were placed by an expert radiologist. An objective of the study was to demonstrate that single-sequence MRI at multiple timepoints provides sufficient information for CNN models to accurately classify lesions.

Yurtakkal et al. [ 35 ] utilized DCE images of 98 benign and 102 malignant lesions, producing 0.98 accuracy, 1.00 sensitivity, and 0.96 specificity. The multi-layer CNN architecture utilized consisted of six groups of convolutional, batch normalization, rectified linear activation function layers, and five max-pooling followed by one dropout layer, one fully connected layer, and one softmax layer.

Zheng et al. [ 36 ] used a dense convolutional long short-term memory (DC-LSTM) on a dataset of lesions obtained through a university hospital ( N  = 72). The method was inspired by DenseNet and built on convolutional LSTM. It first uses a three-layer convolutional LSTM to encode DCE-MRI as sequential data and extract time-intensity information then adds a simplified dense block to reduce the amount of information being processed and improve feature reuse. This lowered the variance and improved accuracy in the results. Compared to a ResNet-50 model trained only on the main task, the combined model of DC-LSTM + ResNet improved the accuracy from 0.625 to 0.847 on the same dataset. Additionally, the authors proposed a latent attributes method to efficiently use the information in diagnostic reports and accelerate the convergence of the network.

Jiejie Zhou et al. [ 37 ] evaluated 133 lesions (91 malignant and 62 benign) using ResNET50, which is similar to ResNET18 used by Truhn et al. [ 33 ] and Haarburger et al . [ 26 ]. Their investigation demonstrated that deep learning produced higher accuracy compared to ROI-based and radiomics-based models in distinguishing between benign and malignant lesions. They compared the metrics resulting from using five different bounding boxes. They found that using the tumor alone and smaller bounding boxes yielded the highest AUC of 0.97–0.99. They also found that the inclusion of a small amount of peritumoral tissue improved accuracy compared to smaller boxes that did not include peritumoral tissue (tumor alone boxes) or larger input boxes (that include tissue more remote from peritumoral tissue), with accuracy of 0.91 in the testing dataset. The tumor microenvironment influences tumor growth, and the tumor itself can alter its microenvironment to become more supportive of tumor growth. Therefore, the immediate peritumoral tissue, which would include the tumor microenvironment, was important in guiding the CNN to accurately differentiate between benign and malignant tumors. This dynamic peritumoral ecosystem can be influenced by the tumor directing heterogeneous cells to aggregate and promote angiogenesis, chronic inflammation, tumor growth, and invasion. Recognizing features displayed by biomarkers of the tumor microenvironment may help to identify and grade the aggressiveness of a lesion. This complex interaction between the tumor and its microenvironment may potentially be a predictor of outcomes as well and should be included in DL models that require segmentation. In DL models using whole images without segmentation of any sort, the peritumoral tissue would already be included, which would preclude the need for precise bounding boxes.

Juan Zhou et al. [ 38 ] used 3D deep learning models to classify and localize malignancy from cases ( N  = 1537) of MRIs. The deep 3D densely connected networks were utilized under image-level supervision (weakly supervised). Since 3D weakly supervised approach was not well studied compared to 2D methods, the purpose of this study was to develop a 3D deep learning model that could identify malignant cancer from benign lesions and could localize the cancer. The model configurations of global average pooling (GAP) and global max-pooling (GMP) that were used both achieved over 0.80 accuracy with AUC of 0.856 (GMP) and 0.858 (GAP) which demonstrated the effectiveness of the 3D DenseNet deep learning method in MRI scans to diagnose breast cancer. The model ensemble achieved AUC of 0.859.

Summary of lessons learned

Most studies were single-center studies, but they came from around the world, with the majority coming from the US, Asia, and Europe. All studies except one [ 33 ] were retrospective studies. The sample size of each study ranged from 42 to 690 patients, generally small for DL analysis. Sample sizes for patients with benign and malignant lesions were comparable and were not skewed toward either normal or malignant lesions, suggesting that these datasets were not from high-risk screening patients because high-risk screening dataset would have consisted of very low (i.e., typically < 5%) positive cases.

Image types

Most studies used private datasets as their image source. ISPY-1 data were the only public dataset noted ( https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=20643859 ). Most studies involved DCE data acquisition, but most analysis include only a single post-contrast MRI. For those that used multiple post-contrast MRI dynamics, most fed each dynamic into each separate independent channel, which does not optimally make use of the relationships between imaging dynamics. Some studies used subtraction of post- and pre-contrast or signal enhancement ratio (SER) [ 24 , 32 , 35 ]. Three studies utilized MIP DCE images to minimize computation cost [ 13 , 24 , 25 ]. However, collapsing images by MIP has drawbacks (i.e., collapse enhanced vascular structures into a single plane may be mistaken as cancer). There were only five studies [ 23 , 26 , 28 , 33 , 36 ] that utilized multiparametric data types (i.e., DCE, T2-weighted, and DWI). Although combining multiple types of MRIs should improve performance, this has not been conclusively demonstrated in practice.

Types of DL architectures

RetinaNet and KFLi are optimized for object detection, while VGGNet, InceptionResNet, and AlexNet are designed for image classification (see review [ 16 , 17 , 39 ]). LSTM is used for time-series modeling. DenseNet, on the other hand, can be used for a wide range of tasks, including image classification, object detection, and semantic segmentation. Ensemble methods, which combine multiple models, are useful for boosting the overall performance of a system. U-Net and R-Net are specialized deep learning models for semantic segmentation tasks in medical image analysis. U-Net uses an encoder–decoder architecture to segment images into multiple classes, while R-Net is a residual network that improves the accuracy and efficiency of the segmentation task.

The most used algorithm is CNN or CNN-based. There is no consensus that certain algorithms are better than others. Given the fact that different algorithms were tested on different datasets, it is not possible to conclude that a particular DL architecture performs better than others. Careful comparison of multiple algorithms on the same datasets is needed. Thus, we only discussed potential advantages and disadvantages of each DL architecture. Performance indices could be misleading.

Although each model has its own unique architecture and design principles, most of the above-mentioned methods utilized convolutional layers, pooling layers, activation functions, and regularization techniques (such as dropout and batch normalization) for model optimization. Additionally, the use of pre-trained models and transfer learning has become increasingly popular, allowing leverage of knowledge learned from large datasets such as ImageNet to improve the performance of their models on smaller, specialized datasets. However, the literature on transfer learning in breast cancer MRI detection is limited. A relatively new deep learning method known as transformer has found exciting applications in medical imaging [ 40 , 41 ].

Ground truths

Ground truths were either based on pathology (i.e., benign versus malignant cancer), radiology reports, radiologist annotation (ROI contoured on images), or a bounding box, with reference to pathology or clinical follow-up (i.e., absence of a positive clinical diagnosis). While the gold standard is pathology, imaging or clinical follow-up without adverse change over a prescribed period has been used as empiric evidence of non-malignancy. This is an acceptable form of ground truth.

Only four out of 18 studies provided heatmaps of the regions that the DL algorithms consider important. Heatmaps enable data to be presented visually in color showing whether the area of activity makes sense anatomically or if it is artifactual (i.e., biopsy clip, motion artifact, or outside of the breast). Heatmaps are important for interpretability and explainability of DL outputs.

Performance

All studies include some performance indices, and most include AUC, accuracy, sensitivity, and specificity. AUC ranged from 0.5 to 1.0, with the majority around 0.8–0.9. Other metrics also varied over a wide range. DL training methods varied, and they included leave-one-out method, hold-out method, and splitting the dataset (such as 80%/20% training/testing) with cross validation. Most studies utilized five- or tenfold cross validation for performance evaluation but some used a single hold-out method, and some did not include cross validation. Cross validation is important to avoid unintentional skewing of data due to partition for training and testing. Different training methods could affect performance. Interpretation of these metrics needs to be made with caution as there could be study reporting bias, small sample size, and overfitting, among others. High-performance indices of the DL algorithm performance are necessary for adoption in clinical use. However, good performance indices alone are not sufficient. Other measures such as heatmaps and experience to gain trust are needed for widespread clinical adoption of DL algorithms.

DL detection of axillary lymph node involvement

Accurate assessment of the axillary lymph node involvement in breast cancer patients is also essential for prognosis and treatment planning [ 42 , 43 ]. Current radiological staging of nodal metastasis has poor accuracy. DL detection of lymph node involvement is challenging because of their small sizes and difficulty in getting ground truths. Only a few studies have reported the use of DL to detect lymph node involvement [ 44 , 45 , 46 ].

Challenges for DL to achieve routine clinical applications

Although deep learning is a promising tool in the diagnosis of breast cancer, there are several challenges that need to be addressed before routine clinical applications can be broadly realized.

Data availability: One of the major challenges in medical image diagnosis (and breast cancer MRI in particular) is the availability of large, diverse, and well-annotated datasets. Deep learning models require a large amount of high-quality data to learn from, but, in many cases, medical datasets are small and imbalanced. In medical image diagnosis, it is important to have high-quality annotations of images, which can be time-consuming and costly to obtain. Annotating medical images requires specialized expertise, and there may be inconsistencies between different experts. This can lead to challenges in building accurate and generalizable models. Medical image datasets can lack diversity, which can lead to biased models. For example, a model trained on images with inadequate representation of racial or ethnicity subgroups may not be broadly generalizable. Private medical datasets obtained from one institution could be non-representative of certain racial or ethnic subgroups and, therefore, may not be generalizable. Publicly available data are unfortunately limited, one of which can be found on cancerimagingarchive.net. Collaborative learning facilitating training of DL models by sharing data without breaching privacy can be accomplished with federated learning [ 47 ].

Interpretability , explainability, and generalizability [ 48 ]: Deep learning models are often seen as “black boxes” that can be difficult to interpret. This is especially problematic in medical image diagnosis, where it is important to understand why a particular diagnosis is made. Recent research has focused on developing methods to explain the decision-making process of deep learning models, such as using attention mechanisms or generating heatmaps to highlight relevant regions in the MRI image. While efforts have been made to develop methods to explain the decision-making process of deep learning models, the explainability of these models is still limited [ 49 ]. This can make it difficult for clinicians to understand the model's decision and to trust the model. Deep learning models may perform well on the datasets on which they were trained but may not generalize well to new datasets or to patients with different characteristics. This can lead to challenges in deploying the model in a real-world setting.

Ethical concerns: Deep learning models can be used to make life-or-death decisions, such as the diagnosis of cancer. This raises ethical concerns about the safety, responsibility, privacy, fairness, and transparency of these models [ 50 ]. There are also social implications (including but not limited to equity) of using artificial intelligence in health care. This needs to be addressed as we develop more and more powerful DL algorithms.

Perspectives and conclusions

Artificial intelligence has the potential to revolutionize breast cancer screening and diagnosis, helping radiologists to be more efficient and more accurate, ultimately leading to better patient outcomes. It can also help to reduce the need for biopsy or unnecessary testing and treatment. However, some challenges exist that preclude broad deployment in clinical practice to date. There need to be large, diverse, and well-annotated images that are readily available for research. Deep learning results need to be more accurate, interpretable, explainable, and generalizable. A future research direction includes incorporation of other clinical data and risk factors into the model, such as age, family history, or genetic mutations, to improve diagnostic accuracy and enable personalized medicine. Another direction is to assess the impact of deep learning on health outcomes to enable more investment in hospital administrators and other stakeholders. Finally, it is important to address the ethical, legal, and social implications of using artificial intelligence.

Availability of data and materials

Not applicable.

Saslow D, Boetes C, Burke W, Harms S, Leach MO, Lehman CD, Morris E, Pisano E, Schnall M, Sener S, et al. American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography. CA Cancer J Clin. 2007;57(2):75–89.

Article   PubMed   Google Scholar  

Feig S. Comparison of costs and benefits of breast cancer screening with mammography, ultrasonography, and MRI. Obstet Gynecol Clin North Am. 2011;38(1):179–96.

Kumar NA, Schnall MD. MR imaging: its current and potential utility in the diagnosis and management of breast cancer. Magn Reson Imaging Clin N Am. 2000;8(4):715–28.

Article   CAS   PubMed   Google Scholar  

Lehman CD, Smith RA. The role of MRI in breast cancer screening. J Natl Compr Canc Netw. 2009;7(10):1109–15.

Mann RM, Kuhl CK, Moy L. Contrast-enhanced MRI for breast cancer screening. J Magn Reson Imaging. 2019;50(2):377–90.

Article   PubMed   PubMed Central   Google Scholar  

Batchu S, Liu F, Amireh A, Waller J, Umair M. A review of applications of machine learning in mammography and future challenges. Oncology. 2021;99(8):483–90.

Wuni AR, Botwe BO, Akudjedu TN. Impact of artificial intelligence on clinical radiography practice: futuristic prospects in a low resource setting. Radiography (Lond). 2021;27(Suppl 1):S69–73.

Skegg D, Paul C, Benson-Cooper D, Chetwynd J, Clarke A, Fitzgerald N, Gray A, St George I, Simpson A. Mammographic screening for breast cancer: prospects for New Zealand. N Z Med J. 1988;101(852):531–3.

CAS   PubMed   Google Scholar  

Wood DA, Kafiabadi S, Busaidi AA, Guilhem E, Montvila A, Lynch J, Townend M, Agarwal S, Mazumder A, Barker GJ, et al. Deep learning models for triaging hospital head MRI examinations. Med Image Anal. 2022;78: 102391.

Mahoro E, Akhloufi MA. Applying deep learning for breast cancer detection in radiology. Curr Oncol. 2022;29(11):8767–93.

Deo RC, Nallamothu BK. Learning about machine learning: the promise and pitfalls of big data and the electronic health record. Circ Cardiovasc Qual Outcomes. 2016;9(6):618–20.

Meyer-Base A, Morra L, Tahmassebi A, Lobbes M, Meyer-Base U, Pinker K. AI-enhanced diagnosis of challenging lesions in breast mri: a methodology and application primer. J Magn Reson Imaging. 2021;54(3):686–702.

Adachi M, Fujioka T, Mori M, Kubota K, Kikuchi Y, Xiaotong W, Oyama J, Kimura K, Oda G, Nakagawa T, et al. Detection and diagnosis of breast cancer using artificial intelligence based assessment of maximum intensity projection dynamic contrast-enhanced magnetic resonance images. Diagnostics (Basel). 2020;10(5):330.

LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–44.

Koh DM, Papanikolaou N, Bick U, Illing R, Kahn CE Jr, Kalpathi-Cramer J, Matos C, Marti-Bonmati L, Miles A, Mun SK, et al. Artificial intelligence and machine learning in cancer imaging. Commun Med (Lond). 2022;2:133.

Oza P, Sharma P, Patel S, Bruno A. A bottom-up review of image analysis methods for suspicious region detection in mammograms. J Imaging. 2021;7(9):190.

Saba T. Recent advancement in cancer detection using machine learning: systematic survey of decades, comparisons and challenges. J Infect Public Health. 2020;13(9):1274–89.

Hu Q, Giger ML. Clinical artificial intelligence applications: breast imaging. Radiol Clin North Am. 2021;59(6):1027–43.

Sechopoulos I, Teuwen J, Mann R. Artificial intelligence for breast cancer detection in mammography and digital breast tomosynthesis: state of the art. Semin Cancer Biol. 2021;72:214–25.

Sheth D, Giger ML. Artificial intelligence in the interpretation of breast cancer on MRI. J Magn Reson Imaging. 2020;51(5):1310–24.

Khan N, Adam R, Huang P, Maldjian T, Duong TQ. Deep learning prediction of pathologic complete response in breast cancer using MRI and other clinical data: a systematic review. Tomography. 2022;8(6):2784–95.

Ayatollahi F, Shokouhi SB, Mann RM, Teuwen J. Automatic breast lesion detection in ultrafast DCE-MRI using deep learning. Med Phys. 2021;48(10):5897–907.

Feng H, Cao J, Wang H, Xie Y, Yang D, Feng J, Chen B. A knowledge-driven feature learning and integration method for breast cancer diagnosis on multi-sequence MRI. Magn Reson Imaging. 2020;69:40–8.

Antropova N, Abe H, Giger ML. Use of clinical MRI maximum intensity projections for improved breast lesion classification with deep convolutional neural networks. J Med Imaging (Bellingham). 2018;5(1): 014503.

PubMed   Google Scholar  

Fujioka T, Yashima Y, Oyama J, Mori M, Kubota K, Katsuta L, Kimura K, Yamaga E, Oda G, Nakagawa T, et al. Deep-learning approach with convolutional neural network for classification of maximum intensity projections of dynamic contrast-enhanced breast magnetic resonance imaging. Magn Reson Imaging. 2021;75:1–8.

Haarburger C, Baumgartner M, Truhn D, Broeckmann M, Schneider H, Schrading S, Kuhl C, Merhof D: Multi scale curriculum CNN for context-aware breast mri malignancy classification. In: Medical Image Computing and Computer Assisted Intervention—MICCAI ; 2019: 495–503.

Herent P, Schmauch B, Jehanno P, Dehaene O, Saillard C, Balleyguier C, Arfi-Rouche J, Jegou S. Detection and characterization of MRI breast lesions using deep learning. Diagn Interv Imaging. 2019;100(4):219–25.

Hu Q, Whitney HM, Giger ML. A deep learning methodology for improved breast cancer diagnosis using multiparametric MRI. Sci Rep. 2020;10(1):10536.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Li J, Fan M, Zhang J, Li L: Discriminating between benign and malignant breast tumors using 3D convolutional neural network in dynamic contrast enhanced-MR images. In: SPIE Medical Imaging : SPIE; 2017: 10138.

Liu MZ, Swintelski C, Sun S, Siddique M, Desperito E, Jambawalikar S, Ha R. Weakly supervised deep learning approach to breast mri assessment. Acad Radiol. 2022;29(Suppl 1):S166–72.

Marrone S, Piantadosi G, Fusco R, Petrillo A, Sansone M, Sansone C: An investigation of deep learning for lesions malignancy classification in breast DCE-MRI. In: International Conference on Image Analysis and Processing (ICIAP) 2017: 479–489.

Rasti R, Teshnehlab M, Phung SL. Breast cancer diagnosis in DCE-MRI using mixture ensemble of convolutional neural networks. Pattern Recogn. 2017;72:381–90.

Article   Google Scholar  

Truhn D, Schrading S, Haarburger C, Schneider H, Merhof D, Kuhl C. Radiomic versus convolutional neural networks analysis for classification of contrast-enhancing lesions at multiparametric breast MRI. Radiology. 2019;290(2):290–7.

Wu W, Wu J, Dou Y, Rubert N, Wang Y. A deep learning fusion model with evidence-based confidence level analysis for differentiation of malignant and benign breast tumors using dynamic contrast enhanced MRI. Biomed Signal Process Control. 2022;72: 103319.

Yurttakal AH, Erbay H, İkizceli T, Karaçavuş S. Detection of breast cancer via deep convolution neural networks using MRI images. Multimed Tools Appl. 2020;79:15555–73.

Zheng H, Gu Y, Qin Y, Huang X, Yang J, Yang G-Z: Small lesion classification in dynamic contrast enhancement MRI for breast cancer early detection. In: International Conference on Medical Image Computing and Computer-Assisted Intervention; 2018.

Zhou J, Zhang Y, Chang KT, Lee KE, Wang O, Li J, Lin Y, Pan Z, Chang P, Chow D, et al. Diagnosis of benign and malignant breast lesions on DCE-MRI by using radiomics and deep learning with consideration of peritumor tissue. J Magn Reson Imaging. 2020;51(3):798–809.

Zhou J, Luo LY, Dou Q, Chen H, Chen C, Li GJ, Jiang ZF, Heng PA. Weakly supervised 3D deep learning for breast cancer classification and localization of the lesions in MR images. J Magn Reson Imaging. 2019;50(4):1144–51.

Alzubaidi L, Zhang J, Humaidi AJ, Al-Dujaili A, Duan Y, Al-Shamma O, Santamaria J, Fadhel MA, Al-Amidie M, Farhan L. Review of deep learning: concepts, CNN architectures, challenges, applications, future directions. J Big Data. 2021;8(1):53.

Li J, Chen J, Tang Y, Wang C, Landman BA, Zhou SK. Transforming medical imaging with transformers? A comparative review of key properties, current progresses, and future perspectives. Med Image Anal. 2023;85: 102762.

Moutik O, Sekkat H, Tigani S, Chehri A, Saadane R, Tchakoucht TA, Paul A. Convolutional neural networks or vision transformers: Who will win the race for action recognitions in visual data? Sensors (Basel). 2023;23(2):734.

Chang JM, Leung JWT, Moy L, Ha SM, Moon WK. Axillary nodal evaluation in breast cancer: state of the art. Radiology. 2020;295(3):500–15.

Zhou P, Wei Y, Chen G, Guo L, Yan D, Wang Y. Axillary lymph node metastasis detection by magnetic resonance imaging in patients with breast cancer: a meta-analysis. Thorac Cancer. 2018;9(8):989–96.

Ren T, Cattell R, Duanmu H, Huang P, Li H, Vanguri R, Liu MZ, Jambawalikar S, Ha R, Wang F, et al. Convolutional neural network detection of axillary lymph node metastasis using standard clinical breast MRI. Clin Breast Cancer. 2020;20(3):e301–8.

Ren T, Lin S, Huang P, Duong TQ. Convolutional neural network of multiparametric MRI accurately detects axillary lymph node metastasis in breast cancer patients with pre neoadjuvant chemotherapy. Clin Breast Cancer. 2022;22(2):170–7.

Golden JA. Deep learning algorithms for detection of lymph node metastases from breast cancer: helping artificial intelligence be seen. JAMA. 2017;318(22):2184–6.

Gupta S, Kumar S, Chang K, Lu C, Singh P, Kalpathy-Cramer J. Collaborative privacy-preserving approaches for distributed deep learning using multi-institutional data. Radiographics. 2023;43(4): e220107.

Miotto R, Wang F, Wang S, Jiang X, Dudley JT. Deep learning for healthcare: review, opportunities and challenges. Brief Bioinform. 2018;19(6):1236–46.

Holzinger A, Langs G, Denk H, Zatloukal K, Muller H. Causability and explainability of artificial intelligence in medicine. Wiley Interdiscip Rev Data Min Knowl Discov. 2019;9(4): e1312.

Smallman M. Multi scale ethics-why we need to consider the ethics of AI in healthcare at different scales. Sci Eng Ethics. 2022;28(6):63.

Download references

Author information

Authors and affiliations.

Department of Radiology, Albert Einstein College of Medicine and the Montefiore Medical Center, 1300 Morris Park Avenue, Bronx, NY, 10461, USA

Richard Adam, Kevin Dell’Aquila, Laura Hodges, Takouhie Maldjian & Tim Q. Duong

You can also search for this author in PubMed   Google Scholar

Contributions

RA performed literature search, analyzed data, and wrote paper. KD performed literature search, analyzed data, and edited paper. LH analyzed literature and edited paper. TM analyzed literature and edited paper. TQD wrote and edited paper, and supervised.

Corresponding author

Correspondence to Tim Q. Duong .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Adam, R., Dell’Aquila, K., Hodges, L. et al. Deep learning applications to breast cancer detection by magnetic resonance imaging: a literature review. Breast Cancer Res 25 , 87 (2023). https://doi.org/10.1186/s13058-023-01687-4

Download citation

Received : 09 March 2023

Accepted : 11 July 2023

Published : 24 July 2023

DOI : https://doi.org/10.1186/s13058-023-01687-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Machine learning
• Artificial intelligence
• Texture feature analysis
• Convolutional neural network
• Dynamic contrast enhancement

Breast Cancer Research

ISSN: 1465-542X

• Submission enquiries: [email protected]

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Recommended Articles
• PubMed/Medline
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

An update of the possible applications of magnetic resonance imaging (mri) in dentistry: a literature review.

1. Introduction

• Scanning artifact. These artifacts are defined by pixels that do not faithfully represent the tissue components studied. The shape of these artifacts depends on the scanning plane, whether it is axial or sagittal. The severity depends on the magnetic properties and position of the present metal; its orientation, shape, number; the homogeneity of the alloy; and the MRI sequence used. On this topic, the literature contains contradictory results, depending on where the attention has been focused on, whether that was gold content alloys, titanium, or a dental amalgam [ 9 , 10 , 11 , 12 ]. Distortion of the static magnetic field is generated from the difference in the magnetic susceptibility, as signal incoherence is generated by substances with different magnetic capacities. In addition to this typology, there are also artifacts caused by eddy currents, induced by alternating gradients and radiofrequency magnetic fields, which participate in generating distortions.
• Mechanical effects (magnetically induced displacement). The most immediate risk associated with the MR environment is the attraction between the MRI device (a magnet) and ferromagnetic metal objects. The magnetic field is strong enough to pull heavy objects towards the scanner at a very high velocity, this is also known as “the projectile effect”. Patients at the highest risk are those with metals not belonging to medical devices (e.g., projectiles, piercings, welding droplets), and among patients with medical devices, those with pacemakers, cochlear implants, neurostimulators, and infusion pumps are at risk. The complications related to RMI can cause malfunction, dislocation, and soft tissue burns (due to the absorption of radiofrequency energy).
• Physical effects (radiofrequency heating). Metallic objects in the human body, such as pacemakers, cochlear implants, neurostimulators, and infusion pumps, before human tissues themselves, can undergo radiofrequency-induced heating. In addition, the batteries of medical devices can also be subject to rapid discharge. Unwanted effects and the mechanisms that generate them are shown in Figure 1 .

2. Materials and Methods

4. discussion, 4.1. fundamental parameters in mri, 4.2. apical periodontitis diagnosis, 4.3. evaluation of dental fractures, 4.4. endodontics, endodontic anatomy and conservative dentistry, 4.5. implantology, 4.6. maxillary sinus diagnosis and surgery, 5. conclusions, author contributions, institutional review board statement, informed consent statement, data availability statement, conflicts of interest.

• Valenti-Obino, F.; di Nardo, D.; Quero, L.; Miccoli, G.; Gambarini, G.; Testarelli, L.; Galli, M. Symmetry of root and root canal morphology of mandibular incisors: A cone-beam computed tomography study in vivo. J. Clin. Exp. Dent. 2019 , 11 , e527–e533. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Pauwels, R.; Araki, K.; Siewerdsen, J.H.; Thongvigitmanee, S.S. Technical aspects of dental CBCT: State of the art. Dentomaxillofacial Radiol. 2015 , 44 , 20140224. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Chogle, S.; Zuaitar, M.; Sarkis, R.; Saadoun, M.; Mecham, A.; Zhao, Y. The Recommendation of Cone-beam Computed Tomography and Its Effect on Endodontic Diagnosis and Treatment Planning. J. Endod. 2020 , 46 , 162–168. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• De Felice, F.; di Carlo, G.; Saccucci, M.; Tombolini, V.; Polimeni, A. Dental Cone Beam Computed Tomography in Children: Clinical Effectiveness and Cancer Risk due to Radiation Exposure. Oncology 2019 , 96 , 173–178. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Di Nardo, D.; Gambarini, G.; Capuani, S.; Testarelli, L. Nuclear Magnetic Resonance Imaging in Endodontics: A Review. J. Endod. 2018 , 44 , 536–542. [ Google Scholar ] [ CrossRef ]
• Piancino, M.G.; Cirillo, S.; Frongia, G.; Cena, F.; Bracco, A.A.; Dalmasso, P.; Bracco, P. Sensitivity of magnetic resonance imaging and computed axiography in the diagnosis of temporomandibular joint disorders in a selected patient population. Int. J. Prosthodont. 2012 , 25 , 120–126. [ Google Scholar ] [ PubMed ]
• Fu, K.Y.; Hu, M.; Yu, Q.; Yang, C.; Cheng, Y.; Long, X.; Zhang, Z.G.; Liu, H.C. Experts consensus on MRI examination specification and diagnostic criteria of temporomandibular joint disc displacement. Chin. J. Stomatol. 2020 , 55 , 608–612. [ Google Scholar ]
• Villanueva-Meyer, J.E.; Mabray, M.C.; Cha, S. Current Clinical Brain Tumor Imaging. Neurosurgery 2017 , 81 , 397–415. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Chockattu, S.J.; Suryakant, D.B.; Thakur, S. Unwanted effects due to interactions between dental materials and magnetic resonance imaging: A review of the literature. Restor. Dent. Endod. 2018 , 43 , e39. [ Google Scholar ] [ CrossRef ]
• Hargreaves, B.A.; Worters, P.W.; Pauly, K.B.; Pauly, J.M.; Koch, K.M.; Gold, G.E. Metal-Induced Artifacts in MRI. Am. J. Roentgenol. 2011 , 197 , 547–555. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Stradiotti, P.; Curti, A.; Castellazzi, G.; Zerbi, A. Metal-related artifacts in instrumented spine. Techniques for reducing artifacts in CT and MRI: State of the art. Eur. Spine J. 2009 , 18 , 102–108. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Hilgenfeld, T.; Prager, M.; Schwindling, F.S.; Jende, J.M.; Rammelsberg, P.; Bendszus, M.; Heiland, S.; Juerchott, A. Protocol for the Evaluation of MRI Artifacts Caused by Metal Implants to Assess the Suitability of Implants and the Vulnerability of Pulse Sequences. J. Vis. Exp. 2018 , 17 , e57394. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Kajima, Y.; Takaichi, A.; Tsutsumi, Y.; Hanawa, T.; Wakabayashi, N.; Kawasaki, A. Influence of magnetic susceptibility and volume on MRI artifacts produced by low magnetic susceptibility Zr-14Nb alloy and dental alloys. Dent. Mater. J. 2020 , 39 , 256–261. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Prager, M.; Heiland, S.; Gareis, D.; Hilgenfeld, T.; Bendszus, M.; Gaudino, C. Dental MRI using a dedicated RF-coil at 3 Tesla. J. Cranio Maxillofac. Surg. 2015 , 43 , 2175–2182. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Klinke, T.; Daboul, A.; Maron, J.; Gredes, T.; Puls, R.; Jaghsi, A.; Biffar, R. Artifacts in Magnetic Resonance Imaging and Computed Tomography Caused by Dental Materials. PLoS ONE 2012 , 7 , e31766. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Roser, C.; Hilgenfeld, T.; Sen, S.; Badrow, T.; Zingler, S.; Heiland, S.; Bendszus, M.; Lux, C.J.; Juerchott, A. Evaluation of magnetic resonance imaging artifacts caused by fixed orthodontic CAD/CAM retainers—An in vitro study. Clin. Oral Investig. 2021 , 25 , 1423–1431. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Geibel, M.-A.; Gelißen, B.; Bracher, A.-K.; Rasche, V. Artifact Properties of Dental Ceramic and Titanium Implants in MRI. RöFo Fortschr. Geb. Röntgenstrahlen Bildgeb. Verfahr. 2018 , 191 , 433–441. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Tymofiyeva, O.; Vaegler, S.; Rottner, K.; Boldt, J.; Hopfgartner, A.J.; Proff, P.C.; Richter, E.-J.; Jakob, P.M. Influence of dental materials on dental MRI. Dentomaxillofacial Radiol. 2013 , 42 , 20120271. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Hilgenfeld, T.; Prager, M.; Heil, A.; Schwindling, F.S.; Nittka, M.; Grodzki, D.; Rammelsberg, P.; Bendszus, M.; Heiland, S. PETRA, MSVAT-SPACE and SEMAC sequences for metal artefact reduction in dental MR imaging. Eur. Radiol. 2017 , 27 , 5104–5112. [ Google Scholar ] [ CrossRef ]
• Duttenhoefer, F.; Mertens, M.E.; Vizkelety, J.; Gremse, F.; Stadelmann, V.A.; Sauerbier, S. Magnetic resonance imaging in zirconia-based dental implantology. Clin. Oral Implant. Res. 2015 , 26 , 1195–1202. [ Google Scholar ] [ CrossRef ]
• Moher, D.; Liberati, A.; Tetzlaff, J.; Altman, D.G.; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. BMJ 2009 , 21 , 339:b2535. [ Google Scholar ]
• Idiyatullin, D.; Corum, C.A.; Nixdorf, D.R.; Garwood, M. Intraoral approach for imaging teeth using the transverse B 1 field components of an occlusally oriented loop coil. Magn. Reson. Med. 2014 , 72 , 160–165. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Siqueira, J.F., Jr. Aetiology of root canal treatment failure: Why well-treated teeth can fail. Int. Endod. J. 2001 , 34 , 1–10. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Nair, P. Pathogenesis of Apical Periodontitis and the Causes of Endodontic Failures. Crit. Rev. Oral Biol. Med. 2004 , 15 , 348–381. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Chang, L.; Umorin, M.; Augsburger, R.A.; Glickman, G.N.; Jalali, P. Periradicular Lesions in Cancellous Bone Can Be Detected Radiographically. J. Endod. 2020 , 46 , 496–501. [ Google Scholar ] [ CrossRef ]
• Siqueira, J.F., Jr.; Rôças, I.N.; Ricucci, D.; Hülsmann, M. Causes and management of post-treatment apical periodontitis. Br. Dent. J. 2014 , 216 , 305–312. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Lizio, G.; Salizzoni, E.; Coe, M.; Gatto, M.R.; Asioli, S.; Balbi, T.; Pelliccioni, G.A. Differential diagnosis between a granuloma and radicular cyst: Effectiveness of magnetic resonance imaging. Int. Endod. J. 2018 , 51 , 1077–1087. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Ricucci, D.; Mannocci, F.; Ford, T.R. A study of periapical lesions correlating the presence of a radiopaque lamina with his-tological findings. Oral Surg. Oral Med. Oral Pathol Oral Radiol. Endod. 2006 , 101 , 389–394. [ Google Scholar ] [ CrossRef ]
• Ricucci, D.; Rôças, I.N.; Hernández, S.; Siqueira, J.F., Jr. “True” Versus “Bay” Apical Cysts: Clinical, Radiographic, Histopathologic, and Histobacteriologic Features. J. Endod. 2020 , 46 , 1217–1227. [ Google Scholar ] [ CrossRef ]
• Juerchott, A.; Pfefferle, T.; Flechtenmacher, C.; Mente, J.; Bendszus, M.; Heiland, S.; Hilgenfeld, T. Differentiation of periapical granulomas and cysts by using dental MRI: A pilot study. Int. J. Oral Sci. 2018 , 10 , 1–8. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Lin, L.M.; Ricucci, D.; Lin, J.; Rosenberg, P.A. Nonsurgical Root Canal Therapy of Large Cyst-like Inflammatory Periapical Lesions and Inflammatory Apical Cysts. J. Endod. 2009 , 35 , 607–615. [ Google Scholar ] [ CrossRef ]
• Geibel, M.A.; Schreiber, E.S.; Bracher, A.K.; Hell, E.; Ulrici, J.; Sailer, L.K.; Ozpeynirci, Y.; Rasche, V. Assessment of Apical Periodontitis by MRI: A Feasibility Study. RöFo Fortschr. Geb. Röntgenstrahlen Bildgeb. Verfahr. 2015 , 187 , 269–275. [ Google Scholar ] [ CrossRef ] [ Green Version ]
• Schestatsky, R.; Dartora, G.; Felberg, R.; Spazzin, A.O.; Sarkis-Onofre, R.; Bacchi, A.; Pereira, G.K.R. Do endodontic retreatment techniques influence the fracture strength of endodontically treated teeth? A systematic review and meta-analysis. J. Mech. Behav. Biomed. Mater. 2019 , 90 , 306–312. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Schuurmans, T.J.; Nixdorf, D.R.; Idiyatullin, D.S.; Law, A.S.; Barsness, B.D.; Roach, S.H.; Gaalaas, L. Accuracy and Reliability of Root Crack and Fracture Detection in Teeth Using Magnetic Resonance Imaging. J. Endod. 2019 , 45 , 750–755.e2. [ Google Scholar ] [ CrossRef ]
• Seracchiani, M.; Miccoli, G.; Di Nardo, D.; Zanza, A.; Cantore, M.; Gambarini, G.; Testarelli, L. Effect of Flexural Stress on Torsional Resistance of NiTi Instruments. J. Endod. 2021 , 47 , 472–476. [ Google Scholar ] [ CrossRef ]
• Gambarini, G.; Miccoli, G.; D’Angelo, M.; Seracchiani, M.; Obino, F.V.; Reda, R.; Testarelli, L. The relevance of operative torque and torsional resistance of nickel-titanium rotary instruments: A preliminary clinical investigation. Saudi Endod. J. 2020 , 10 , 260–264. [ Google Scholar ]
• Mazzoni, A.; Pacifici, A.; Zanza, A.; Giudice, A.; Reda, R.; Testarelli, L.; Gambarini, G.; Pacifici, L. Assessment of Real-Time Operative Torque during Nickel-Titanium Instrumentation with Different Lubricants. Appl. Sci. 2020 , 10 , 6201. [ Google Scholar ] [ CrossRef ]
• Gambarini, G.; Seracchiani, M.; Zanza, A.; Miccoli, G.; Del Giudice, A.; Testarelli, L. Influence of shaft length on torsional behavior of endodontic nickel-titanium instruments. Odontology 2020 , 1–6. [ Google Scholar ] [ CrossRef ]
• Tanasiewicz, M. Magnetic Resonance Imaging in Endodontic Treatment Prediction. J. Med. Imaging Radiat. Sci. 2010 , 41 , 127–132. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Bracher, A.-K.; Hofmann, C.; Bornstedt, A.; Hell, E.; Janke, F.; Ulrici, J.; Haller, B.; Geibel, M.-A.; Rasche, V. Ultrashort echo time (UTE) MRI for the assessment of caries lesions. Dentomaxillofacial Radiol. 2013 , 42 , 20120321. [ Google Scholar ] [ CrossRef ]
• Lloyd, C.; Scrimgeour, S.; Chudek, J.; Hunter, G.; Mackay, R. Application of Magnetic Resonance Microimaging to the Study of Dental Caries. Caries Res. 1999 , 34 , 53–58. [ Google Scholar ] [ CrossRef ]
• Ploder, O.; Partik, B.; Rand, T.; Fock, N.; Voracek, M.; Undt, G.; Baumann, A. Reperfusion of autotransplanted teeth—comparison of clinical measurements by means of dental magnetic resonance imaging. Oral Surg. Oral Med. Oral Pathol. Oral Radiol. Endodontology 2001 , 92 , 335–340. [ Google Scholar ] [ CrossRef ]
• Nakashima, M.; Akamine, A. The Application of Tissue Engineering to Regeneration of Pulp and Dentin in Endodontics. J. Endod. 2005 , 31 , 711–718. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Annibali, S.; Bellavia, D.; Ottolenghi, L.; Cicconetti, A.; Cristalli, M.P.; Quaranta, R.; Pilloni, A. Micro-CT and PET analysis of bone regeneration induced by biodegradable scaffolds as carriers for dental pulp stem cells in a rat model of calvarial “critical size” defect: Preliminary data. J. Biomed. Mater. Res. Part. B Appl. Biomater. 2013 , 102 , 815–825. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Assaf, A.T.; Zrnc, T.A.; Remus, C.C.; Khokale, A.; Habermann, C.R.; Schulze, D.; Fiehler, J.; Heiland, M.; Sedlacik, J.; Friedrich, R.E. Early detection of pulp necrosis and dental vitality after traumatic dental injuries in children and adolescents by 3-Tesla magnetic resonance imaging. J. Cranio Maxillofac. Surg. 2015 , 43 , 1088–1093. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Sedlacik, J.; Kutzner, D.; Khokale, A.; Schulze, D.; Fiehler, J.; Celik, T.; Gareis, D.; Smeets, R.; Friedrich, R.E.; Heiland, M.; et al. Optimized 14 + 1 receive coil array and position system for 3D high-resolution MRI of dental and maxillomandibular structures. Dentomaxillofacial Radiol. 2016 , 45 , 20150177. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Seracchiani, M.; Miccoli, G.; Reda, R.; Zanza, A.; Valenti Obino, F.; Bhandi, S.; Gambarini, G.; Testarelli, L. A Comprehensive In Vitro Comparison of Mechanical Properties of Two Rotary Endodontic Instruments. World J. Dent. 2020 , 11 , 185–188. [ Google Scholar ]
• Probst, F.A.; Schweiger, J.; Stumbaum, M.J.; Karampinos, D.; Burian, E.; Probst, M. Magnetic resonance imaging based comput-er-guided dental implant surgery—A clinical pilot study. Clin. Implant. Dent. Relat. Res. 2020 , 22 , 612–621. [ Google Scholar ] [ CrossRef ]
• Gambarini, G.; Plotino, G.; Grande, N.M.; Testarelli, L.; Prencipe, M.; Messineo, D.; Fratini, L.; D’Ambrosio, F. Differential diagnosis of endodontic-related inferior alveolar nerve paraesthesia with cone beam computed tomography: A case report. Int. Endod. J. 2010 , 44 , 176–181. [ Google Scholar ] [ CrossRef ]
• Di Bari, R.; Coronelli, R.; Cicconetti, A. Intraosseous vascularization of anterior mandible: A radiographic analysis. J. Craniofac. Surg. 2014 , 25 , 872–879. [ Google Scholar ] [ CrossRef ]
• Guarnieri, R.; Miccoli, G.; Reda, R.; Mazzoni, A.; Di Nardo, D.; Testarelli, L. Laser microgrooved vs. machined healing abutment disconnection/reconnection: A comparative clinical, radiographical and biochemical study with split-mouth design. Int. J. Implant. Dent. 2021 , 7 , 1–11. [ Google Scholar ] [ CrossRef ]
• Hilgenfeld, T.; Kästel, T.; Heil, A.; Rammelsberg, P.; Heiland, S.; Bendszus, M.; Schwindling, F.S. High-resolution dental magnetic resonance imaging for planning palatal graft surgery—A clinical pilot study. J. Clin. Periodontol. 2018 , 45 , 462–470. [ Google Scholar ] [ CrossRef ]
• Aktuna Belgin, C.; Colak, M.; Adiguzel, O.; Akkus, Z.; Orhan, K. Three-dimensional evaluation of maxillary sinus volume in different age and sex groups using CBCT. Eur. Arch. Otorhinolaryngol. 2019 , 276 , 1493–1499. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Dong, Y.; Zhou, B.; Wang, C.-S.; Huang, Q.; Cui, S.-J.; Li, Y.-C.; Wang, X.-Y. CT and MRI diagnosis of lesions in unilateral maxillary sinus. Chin. J. Otorhinolaryngol. Head Neck Surg. 2013 , 48 , 895–900. [ Google Scholar ]
• Panou, E.; Motro, M.; Ateş, M.; Acar, A.; Erverdi, N. Dimensional changes of maxillary sinuses and pharyngeal airway in Class III patients undergoing bimaxillary orthognathic surgery. Angle Orthod. 2013 , 83 , 824–831. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Özdemir, M.; Kavak, R.P. Season, Age and Sex-Related Differences in Incidental Magnetic Resonance Imaging Findings of Paranasal Sinuses in Adults. Turk. Arch. Otorhinolaryngol. 2019 , 57 , 61–67. [ Google Scholar ] [ CrossRef ]
• Rani, S.U.; Rao, G.V.; Kumar, D.R.; Sravya, T.; Sivaranjani, Y.; Kumar, M.P. Age and gender assessment through three-dimensional morphometric analysis of maxillary sinus using magnetic resonance imaging. J. Forensic Dent. Sci. 2017 , 9 , 46. [ Google Scholar ] [ PubMed ]
• Butaric, L.N.; Wadle, M.; Gascon, J. Pediatric Imaging, Neurocognition and Genetics Study. Anatomical Variation in Maxillary Sinus Ostium Positioning: Implications for Nasal-Sinus Disease. Anat. Rec. 2019 , 302 , 917–930. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Yuan, X.-P.; Li, C.-X.; Cao, Y.; Singh, S.; Zhong, R. Inflammatory myofibroblastic tumour of the maxillary sinus: CT and MRI findings. Clin. Radiol. 2012 , 67 , e53–e57. [ Google Scholar ] [ CrossRef ]
• Laurino, F.A.R.; Gil Choi, I.G.; Kim, J.H.; Gialain, I.O.; Ferraço, R.; Haetinger, R.G.; Pinhata-Baptista, O.H.; Abdala-Junior, R.; Costa, C.; Cortes, A.R.G. Correlation between magnetic resonance imaging and cone-beam computed tomography for maxillary sinus graft assessment. Imaging Sci. Dent. 2020 , 50 , 93–98. [ Google Scholar ] [ CrossRef ]
• Munhoz, L.; Júnior, R.A.; Arita, E.S. The value of the apparent diffusion coefficient calculated from diffusion-weighted magnetic resonance imaging scans in the differentiation of maxillary sinus inflammatory diseases. Oral Surg. Oral Med. Oral Pathol. Oral Radiol. 2019 , 127 , 433–443. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Wang, L.; Chen, K.C.; Gao, Y.; Shi, F.; Liao, S.; Li, G.; Shen, S.G.F.; Yan, J.; Lee, P.K.M.; Chow, B.; et al. Automated bone segmentation from dental CBCT images using patch-based sparse representation and convex optimization. Med. Phys. 2014 , 41 , 043503. [ Google Scholar ] [ CrossRef ] [ PubMed ] [ Green Version ]
• Liu, W.; Wang, Y.; Jiang, T.; Chi, Y.; Zhang, L.; Hua, X.S. Landmarks Detection with Anatomical Constraints for Total Hip Arthroplasty Preoperative Measurements. In Part I, Proceedings of the Medical Image Computing and Computer Assisted Intervention—MICCAI 2020, Lima, Peru, 4–8 October 2020 ; Martel, A.L., Abolmaesumi, P., Stoyanov, D., Mateus, D., Zuluaga, M.A., Zhou, S.K., Racoceanu, D., Joskowicz, L., Eds.; Springer: Cham, Switzerland, 2020; Volume 12264. [ Google Scholar ]
• Zhong, Z.; Li, J.; Zhang, Z.; Jiao, Z.; Gao, X. An Attention-Guided Deep Regression Model for Landmark Detection in Cephalograms. In Proceedings of the Languages and Compilers for Parallel Computing, Shenzhen, China, 13–17 October 2019 ; Springer Science and Business Media LLC: Cham, Switzerland, 2019; pp. 540–548. [ Google Scholar ]

Click here to enlarge figure

Share and Cite

Reda, R.; Zanza, A.; Mazzoni, A.; Cicconetti, A.; Testarelli, L.; Di Nardo, D. An Update of the Possible Applications of Magnetic Resonance Imaging (MRI) in Dentistry: A Literature Review. J. Imaging 2021 , 7 , 75. https://doi.org/10.3390/jimaging7050075

Reda R, Zanza A, Mazzoni A, Cicconetti A, Testarelli L, Di Nardo D. An Update of the Possible Applications of Magnetic Resonance Imaging (MRI) in Dentistry: A Literature Review. Journal of Imaging . 2021; 7(5):75. https://doi.org/10.3390/jimaging7050075

Reda, Rodolfo, Alessio Zanza, Alessandro Mazzoni, Andrea Cicconetti, Luca Testarelli, and Dario Di Nardo. 2021. "An Update of the Possible Applications of Magnetic Resonance Imaging (MRI) in Dentistry: A Literature Review" Journal of Imaging 7, no. 5: 75. https://doi.org/10.3390/jimaging7050075

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

Log in using your username and password

• Search More Search for this keyword Advanced search
• Latest content
• Current issue
• BMJ Journals

You are here

• Volume 75, Issue 9
• A review of structural magnetic resonance neuroimaging
• Article Text
• Article info
• Citation Tools
• Rapid Responses
• Article metrics

• M Symms 1 ,
• H R Jäger 2 ,
• K Schmierer 3 ,
• T A Yousry 2
• 1 Department of Clinical and Experimental Epilepsy, Institute of Neurology, London, UK
• 2 Lysholm Department of Neuroradiology, National Hospital of Neurology and Neurosurgery, Institute of Neurology, London, UK
• 3 Department of Neuroinflammation, NMR Research Unit, Institute of Neurology, London, UK; and Klinik für Neurologie, Chasité, Humbeldt Universitãt zu Berlin, Berlin, Germany
• Correspondence to:
 Prof T A Yousry
 Lysholm Department of Neuroradiology, National Hospital of Neurology and Neurosurgery, Institute of Neurology, Queen Square, London WC1N3BG, UK; t.yousryion.ucl.ac.uk

Magnetic resonance imaging (MRI) is often divided into structural MRI and functional MRI (fMRI). The former is a widely used imaging technique in research as well as in clinical practice. This review describes the more important developments in structural MRI in recent years, including high resolution imaging, T2 relaxation measurement, T2*-weighted imaging, T1 relaxation measurement, magnetisation transfer imaging, and diffusion imaging. The principles underlying these techniques, as well as their use in research and in clinical practice, will be discussed.

• ADC, apparent diffusion coefficient
• AD, Alzheimer’s disease
• CSF, cerebrospinal fluid
• CT, computed tomography
• DTI, diffusion tensor imaging
• DWI, diffusion-weighted imaging
• FLAIR, fluid attenuated inversion recovery
• fMRI, functional magnetic resonance imaging
• Gd, gadolinium
• GRE, gradient echo
• MD, mean diffusivity
• MS, multiple sclerosis
• MTI, magnetisation transfer imaging
• MTR, magnetisation transfer ratio
• NMR, nuclear magnetic resonance
• PP/SPMS, primary progressive/secondary progressive MS
• RF, radiofrequency
• RRMS, relapsing remitting MS
• SNR, signal to noise ratio
• magnetic resonance neuroimaging

https://doi.org/10.1136/jnnp.2003.032714

Statistics from Altmetric.com

Request permissions.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Competing interests: none declared

Read the full text or download the PDF:

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

Quality assurance of human functional magnetic resonance imaging: a literature review

Affiliations.

• 1 Medical Engineering and Technical Center, Shandong First Medical University & Shandong Academy of Medical Sciences, Tai'an 271016, China.
• 2 Department of Radiology, Shandong First Medical University & Shandong Academy of Medical Sciences, Tai'an 271016, China.
• PMID: 31367569
• PMCID: PMC6629553
• DOI: 10.21037/qims.2019.04.18

Functional magnetic resonance imaging (fMRI) has been a popular approach in brain research over the past 20 years. It offers a noninvasive method to probe the brain and uses blood oxygenation level dependent (BOLD) signal changes to access brain function. However, the BOLD signal only represents a small fraction of the total MR signal. System instability and various noise have a strong impact on the BOLD signal. Additionally, fMRI applies fast imaging technique to record brain cognitive process over time, requiring high temporal stability of MR scanners. Furthermore, data acquisition, image quality, processing, and statistical analysis methods also have a great effect on the results of fMRI studies. Quality assurance (QA) programs for fMRI can test the stability of MR scanners, evaluate the quality of fMRI and help to find errors during fMRI scanning, thereby greatly enhancing the success rate of fMRI. In this review, we focus on previous studies which developed QA programs and methods in SCI/SCIE citation peer-reviewed publications over the last 20 years, including topics on existing fMRI QA programs, QA phantoms, image QA metrics, quality evaluation of existing preprocessing pipelines and fMRI statistical analysis methods. The summarized studies were classified into four categories: QA of fMRI systems, QA of fMRI data, quality evaluation of data processing pipelines and statistical methods and QA of task-related fMRI. Summary tables and figures of QA programs and metrics have been developed based on the comprehensive review of the literature.

Keywords: Functional magnetic resonance imaging (fMRI); phantom; quality assurance (QA); quality check.

PubMed Disclaimer

Conflict of interest statement

Conflicts of Interest: The authors have no conflicts of interest to declare.

Two types of functional magnetic…

Two types of functional magnetic resonance imaging (fMRI) quality assurance (QA) phantoms and…

Quality check of the fMRI…

Quality check of the fMRI data. (A) Quality assurance (QA) procedures of the…

Similar articles

• Design of a fused phantom for quantitative evaluation of brain metabolites and enhanced quality assurance testing for magnetic resonance imaging and spectroscopy. Song KH, Kim SY, Lee DW, Jung JY, Lee JH, Baek HM, Choe BY. Song KH, et al. J Neurosci Methods. 2015 Nov 30;255:75-84. doi: 10.1016/j.jneumeth.2015.08.005. Epub 2015 Aug 13. J Neurosci Methods. 2015. PMID: 26277420
• The Marburg-Münster Affective Disorders Cohort Study (MACS): A quality assurance protocol for MR neuroimaging data. Vogelbacher C, Möbius TWD, Sommer J, Schuster V, Dannlowski U, Kircher T, Dempfle A, Jansen A, Bopp MHA. Vogelbacher C, et al. Neuroimage. 2018 May 15;172:450-460. doi: 10.1016/j.neuroimage.2018.01.079. Epub 2018 Feb 1. Neuroimage. 2018. PMID: 29410079
• Resting state fMRI scanner instabilities revealed by longitudinal phantom scans in a multi-center study. Kayvanrad A, Arnott SR, Churchill N, Hassel S, Chemparathy A, Dong F, Zamyadi M, Gee T, Bartha R, Black SE, Lawrence-Dewar JM, Scott CJM, Symons S, Davis AD, Hall GB, Harris J, Lobaugh NJ, MacQueen G, Woo C, Strother S; ONDRI Founding Investigators; CAN-BIND Investigators. Kayvanrad A, et al. Neuroimage. 2021 Aug 15;237:118197. doi: 10.1016/j.neuroimage.2021.118197. Epub 2021 May 21. Neuroimage. 2021. PMID: 34029737
• Report on a multicenter fMRI quality assurance protocol. Friedman L, Glover GH. Friedman L, et al. J Magn Reson Imaging. 2006 Jun;23(6):827-39. doi: 10.1002/jmri.20583. J Magn Reson Imaging. 2006. PMID: 16649196 Review.
• Functional magnetic resonance imaging: the basics of blood-oxygen-level dependent (BOLD) imaging. Forster BB, MacKay AL, Whittall KP, Kiehl KA, Smith AM, Hare RD, Liddle PF. Forster BB, et al. Can Assoc Radiol J. 1998 Oct;49(5):320-9. Can Assoc Radiol J. 1998. PMID: 9803232 Review.
• Fast 3D fMRI acquisition with high spatial resolutions over a reduced FOV. Luo Q, Sun K, Dan G, Zhou XJ. Luo Q, et al. Magn Reson Med. 2024 Nov;92(5):1952-1964. doi: 10.1002/mrm.30191. Epub 2024 Jun 18. Magn Reson Med. 2024. PMID: 38888135
• A functional MRI pre-processing and quality control protocol based on statistical parametric mapping (SPM) and MATLAB. Di X, Biswal BB. Di X, et al. Front Neuroimaging. 2023 Jan 10;1:1070151. doi: 10.3389/fnimg.2022.1070151. eCollection 2022. Front Neuroimaging. 2023. PMID: 37555150 Free PMC article.
• An implementation of integrated information theory in resting-state fMRI. Nemirovsky IE, Popiel NJM, Rudas J, Caius M, Naci L, Schiff ND, Owen AM, Soddu A. Nemirovsky IE, et al. Commun Biol. 2023 Jul 5;6(1):692. doi: 10.1038/s42003-023-05063-y. Commun Biol. 2023. PMID: 37407655 Free PMC article.
• Classification of Migraine Using Static Functional Connectivity Strength and Dynamic Functional Connectome Patterns: A Resting-State fMRI Study. Nie W, Zeng W, Yang J, Zhao L, Shi Y. Nie W, et al. Brain Sci. 2023 Mar 31;13(4):596. doi: 10.3390/brainsci13040596. Brain Sci. 2023. PMID: 37190561 Free PMC article.
• Functional connectivity MRI quality control procedures in CONN. Morfini F, Whitfield-Gabrieli S, Nieto-Castañón A. Morfini F, et al. Front Neurosci. 2023 Mar 23;17:1092125. doi: 10.3389/fnins.2023.1092125. eCollection 2023. Front Neurosci. 2023. PMID: 37034165 Free PMC article.
• Smith K. Brain imaging: fMRI 2.0. Nature 2012;484: 24-6. 10.1038/484024a - DOI - PubMed
• Poldrack RA, Farah MJ. Progress and challenges in probing the human brain. Nature 2015;526:371-9. 10.1038/nature15692 - DOI - PubMed
• Eklund A, Nichols TE, Knutsson H. Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates. Proc Natl Acad Sci U S A 2016;113 7900-5. 10.1073/pnas.1602413113 - DOI - PMC - PubMed
• Tovar DA, Zhan W, Rajan SS. A Rotational Cylindrical fMRI Phantom for Image Quality Control. Plos One 2015;10:e0143172. 10.1371/journal.pone.0143172 - DOI - PMC - PubMed
• Olsrud J, Nilsson A, Mannfolk P, Waites A, Ståhlberg F. A two-compartment gel phantom for optimization and quality assurance in clinical BOLD fMRI. Magn Reson Imaging 2008;26:279-86. 10.1016/j.mri.2007.06.010 - DOI - PubMed

Publication types

• Search in MeSH

Related information

Linkout - more resources, full text sources.

• AME Publishing Company
• Europe PubMed Central
• PubMed Central

Miscellaneous

• NCI CPTAC Assay Portal

• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.