HealthcareCase Study

NYU Langone Health: AI Matches Radiologists in Breast Cancer MRI Diagnosis

NYU Langone Health: AI Matches Radiologists in Breast Cancer MRI Diagnosis

·9 min read
NYU Langone HealthAcademic Medical Center21,537 MRI studies · 13,463 patients in training dataset

Key People & Companies

NYU Langone Health

Deep Learning AI System for Breast Cancer MRI

NYU Langone Health · Internal

Krzysztof J. Geras

Assistant Professor, Department of Radiology at NYU Grossman School of Medicine / NYU Langone Health

Jan Witowski

Postdoctoral Fellow, MD, PhD at NYU Langone Health / NYU School of Medicine

Linda Moy

Professor of Radiology at NYU Langone Health

DCE-MRI

Imaging_technology

BI-RADS 4

American College of Radiology · diagnostic_category

3D-ResNet18

NYU Langone Health (custom implementation) · Framework

+ 8 more entities in the full study

Key Results

  • The AI system achieved an AUROC of 0.924 (95% CI: 0.915-0.933) on NYU Langone's internal test set of 3,936 MRI exams, statistically equivalent to the average performance of five board-certified breast radiologists (AUROC 0.890, range 0.850-0.948).
  • External validation demonstrated generalizability: AUROC of 0.788 on 397 studies from Jagiellonian University Hospital in Poland, and 0.969 on 922 studies from Duke University.
  • Hybrid predictions—averaging AI scores with radiologist assessments—improved AUROC by 0.05 and AUPRC by 0.07 compared to radiologists alone, while boosting interreader agreement from a Fleiss' kappa of 0.56 to 0.77.
  • + 3 more results inside

Breast MRI exams are difficult and time-consuming to interpret, even for experienced radiologists. AI has tremendous potential in improving medical diagnosis, as it can learn from tens of thousands of exams. Using AI to assist radiologists can make the process more accurate, and provide a higher level of confidence in the results.

Krzysztof J. Geras, Assistant Professor, Department of Radiology at NYU Grossman School of Medicine

Sign in to unlock the full case study

Get the complete analysis, all results, expert quotes, and entity profiles — free.