Back to publications
Breast MRIClinical ValidationLesion Detection

Assessment of AI Performance in Detecting Breast Lesions on MRI: A Retrospective Study in a Chinese Cohort

Insights into Imaging / ECR / 2025 / conference

Bayer S, Bhandary Panambur A, Du X, Grimm R, Von Busch H, Maier A, Huang J, Lui S

DOI Poster
Example AI interface with segmented breast MRI lesion from ECR 2025 poster
Local visual summary of the breast MRI AI detection workflow described in the ECR EPOS poster. EPOS source

Scientific summary

This retrospective breast MRI study evaluates lesion-level AI performance in an external Chinese cohort, with emphasis on sensitivity and false-positive burden.

Abstract

This ECR 2025 exhibit evaluates an AI model for suspicious breast lesion detection in breast MRI scans from a Chinese cohort.

Why it matters

External cohort validation is essential for assessing whether breast MRI AI systems can support high-volume screening without unacceptable false-positive rates.

Contribution

It reports retrospective performance of a Siemens Healthineers breast MRI research prototype on West China Hospital data.

Method overview

The study sampled 200 cases from a 4,000-patient breast MRI cohort and used multi-sequence 3T MRI with expert lesion annotations.

Key findings

  • The EPOS poster reports case-level sensitivity of 0.906.
  • The poster reports median 0 false positives per case and average 0.635 false positives per case.
  • The cohort came from West China Hospital MRI exams between November 2017 and August 2023.

Citation

Bayer S, Bhandary Panambur A, Du X, Grimm R, Von Busch H, Maier A, Huang J, Lui S. (2025). Assessment of AI Performance in Detecting Breast Lesions on MRI: A Retrospective Study in a Chinese Cohort. Insights into Imaging / ECR.

TODO: Add BibTeX.

Related publications