Scientists create advanced deep learning model for breast cancer prediction

Researchers have made a groundbreaking advancement in the field of breast cancer detection with the development of a new, interpretable artificial intelligence (AI) model. This model, known as AsymMirai, has the ability to predict 5-year breast cancer risk from mammograms, offering a potential game-changer in early detection and treatment.

Breast cancer is a significant health concern, affecting 1 in 8 women in the U.S. throughout their lifetime. Mammography screening has been a crucial tool in detecting breast cancer early, but the challenge lies in accurately predicting which women are at risk through screening alone.

The study, published in Radiology, compared the performance of AsymMirai to the existing AI model, Mirai, which had shown promise in predicting breast cancer risk. However, Mirai’s complex reasoning process raised concerns about potential overreliance and incorrect diagnoses.

Lead author Jon Donnelly, a Ph.D. student at Duke University, explained that AsymMirai offers a simpler and more understandable approach to predicting breast cancer risk. By focusing on tissue differences between the left and right breasts, the new model achieved comparable results to Mirai in 1- to 5-year risk prediction.

The findings not only underscore the importance of breast asymmetry in cancer detection but also highlight the potential of bilateral dissimilarity as a future imaging marker for breast cancer risk. With its transparent reasoning process, AsymMirai could serve as a valuable tool for radiologists in enhancing breast cancer diagnoses and risk prediction.

This innovative AI model has the potential to revolutionize breast cancer screening practices, potentially influencing the frequency of mammograms for women in the near future. The study’s results offer hope for improved early detection and treatment outcomes in the fight against breast cancer.