A longitudinal DBT-based deep learning model outperformed conventional mammography and clinical models for five-year breast cancer risk prediction.
Key Details
- 1Study leveraged 313,335 DBT exams from 161,077 women (2016–2020) at NYU Langone Health.
- 2DBT model's AUC for five-year risk prediction was 0.72 (independent test set), outperforming single-timepoint DBT (0.71), Mirai (0.69), and Tyrer-Cuzick (0.56, case-control cohort).
- 3Model classified 39.7% of extremely dense breasts as average risk (0.8% incidence in 5 years) and 14.8% of fatty breasts as high risk (2.6% incidence).
- 4Enhanced volumetric and longitudinal DBT imaging contributed to improved prediction.
- 5Authors call for external validation across diverse sites and prospective clinical studies.
Why It Matters
This research supports using AI on DBT images for dynamic and individualized breast cancer risk assessment, potentially refining personalized screening strategies and improving early cancer detection strategies in radiology practice.

Source
AuntMinnie
Related News

•AuntMinnie
Deep Learning PET Biomarkers Outperform Clinical Tools in Prostate Cancer Survival Prediction
Deep learning-derived PET biomarkers more accurately predict metastatic prostate cancer survival than standard clinical risk models.

•AuntMinnie
Multinational AI Model Flags Osteoporosis on Chest X-rays
AI can opportunistically detect osteoporosis from chest x-rays across diverse international populations.

•AuntMinnie
LLM-Based Chart Summarization Tool Reduces Radiologist Review Time
An LLM-based chart summarization tool safely reduced radiologist study review time by 17%.