Predicting Precancerous Changes In High-risk Women: A Breakthrough Mammography-based Deep Learning Approach

This study examined the potential of a deep learning (DL) model in detecting precancerous changes in women at high risk for breast cancer. The DL model was trained on an extensive dataset of screening mammograms and its performance was assessed using the area under the receiver operating characteristic curve (AUC). The results showed that the DL model achieved a one-year AUC of 71 percent and a five-year AUC of 65 percent for predicting breast cancer, outperforming the traditional Breast Imaging Reporting and Data System (BI-RADS) system. The study also found that the DL model had the potential to detect early or premalignant changes that may not be apparent in standard mammograms. Lastly, the combination of the DL model’s results with BI-RADS scores demonstrated improved short-term risk stratification.