Researchers Develop Deep Learning Model to Predict Breast Cancer
A new study published in Radiology reveals the development of an interpretable artificial intelligence (AI) model, named AsymMirai, for predicting 5-year breast cancer risk from mammograms. Traditional mammography screening can significantly lower the risk of breast cancer-related mortality, but accurately predicting which women will develop breast cancer remains challenging. While previous AI models, such as Mirai, have shown promise in this regard, their complex reasoning process raises concerns about overreliance and potential misdiagnoses. Lead author Jon Donnelly and colleagues at Duke University developed AsymMirai, a simplified AI method that enhances interpretability compared to Mirai. By analyzing 210,067 mammograms from 81,824 patients, the researchers demonstrated that AsymMirai performs almost as well as Mirai in predicting 1- to 5-year breast cancer risk. AsymMirai's straightforward reasoning process makes it a valuable tool for assisting radiologists in breast cancer diagnosis and risk prediction.
#ai/software
#clinical decision support