Mammography-based deep learning model may help detect precancerous changes in high-risk women
Emerging research implies a mammography-based deep learning (DL) model may aid in identifying precancerous changes in high-risk women. The study, recently published in 'Radiology: Artificial Intelligence', featured a dataset enriched with African American women, BRCA mutation carriers and those with benign breast disease. The researchers utilised Mirai, a DL model developed by AI faculty lead Regina Barzilay and her colleagues at the MIT Jameel Clinic. In this experiment, Mirai was trained on nearly 211,000 screening mammograms, to assess 6,266 mammography exams from 2,043 women. Findings demonstrate that the DL model achieved a five-year Area Under the Curve (AUC) of 65% for predicting breast cancer, compared to 54% for the Breast Imaging Reporting and Data System (BI-RADS). The researchers also note the significant role of imaging the breast with future cancer within the AI model. Mirroring demonstrated a 62% AUC for positive mirroring and a 51% AUC for negative mirroring. This suggests that DL could detect precancerous or early malignant changes before they become apparent. While the research acknowledges limitations, such as a limited sample size for other racial groups beyond Black and White women, the results suggest that DL tools could enhance the assessment of screening mammograms for more accurate near-term risk stratification.