Related Experiment Video
Updated: Sep 19, 2025
15:48
Tracking the Mammary Architectural Features and Detecting Breast Cancer with Magnetic Resonance Diffusion Tensor Imaging
Published on: December 15, 2014
22.6K
Adaptive Breast MRI Scanning Using AI
Sarah Eskreis-Winkler1, Arka Bhowmik1, Lori H Kelly1
1Department of Radiology, Memorial Sloan Kettering Cancer Center, 300 E 66th St, New York, NY 10065.
Radiology
|June 3, 2025
Summary
Artificial intelligence (AI) can streamline breast MRI screening by directing stratified scanning, reducing scan times without compromising diagnostic accuracy. This AI-directed approach maintains high sensitivity and specificity, ensuring effective cancer detection in abbreviated breast MRI protocols.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Oncology and Cancer Screening
Background:
- Standard MRI protocols for breast cancer screening are often lengthy, posing a challenge for patient throughput and resource allocation.
- The need for efficient yet accurate screening methods is critical in managing the growing demand for breast MRI examinations.
Purpose of the Study:
- To simulate and evaluate the diagnostic performance of artificial intelligence (AI)-directed stratified scanning for breast MRI screening.
- To compare the efficacy of an AI triage system against traditional full breast MRI protocols using various threshold settings.
Main Methods:
- A retrospective reader study analyzed 1423 contrast-enhanced screening breast MRI examinations from three cancer sites (2013-2019).
- An in-house AI tool assigned suspicion scores to identify examinations suitable for an abbreviated breast MRI (AB-MRI) protocol, focusing on dynamic contrast-enhanced MRI scans.
- Diagnostic performance metrics were compared between AI-directed stratified scanning (triage threshold at 50th percentile) and standard full MRI protocols.
Main Results:
- AI-directed stratified scanning demonstrated comparable diagnostic performance to the full MRI protocol, with sensitivity at 88.2% vs 86.3% and specificity at 80.8% vs 81.4%.
- The AI triage resulted in a minimal decrease in specificity (≤2.7 percentage points) while maintaining cancer detection rates and interval cancer rates.
- No additional cancer diagnoses were missed in the AI-triaged examinations that would have been detected by the full MRI protocol.
Conclusions:
- AI-directed stratified MRI scanning offers a viable strategy to reduce simulated examination times for breast MRI screening.
- This AI approach effectively maintains diagnostic performance, including sensitivity, specificity, and cancer detection rates, compared to conventional full protocols.
- The findings support the potential of AI in optimizing breast MRI workflows for increased efficiency without compromising patient care.
