Longitudinal Multi-View Breast Cancer Risk Prediction

The paper introduces LMV-Net, a breast cancer risk prediction model that jointly processes anatomically complementary craniocaudal (CC) and mediolateral oblique (MLO) mammography views within an explicitly aligned longitudinal framework. The authors evaluate it on the public EMBED and CSAW-CC datasets and report consistent improvements over existing breast cancer risk prediction methods, including across breast-density and cancer subgroups. The stated motivation is to combine spatial complementarity across views with temporal information across screening exams. Code is available on GitHub, although the abstract does not provide quantitative metrics, confidence intervals, or details of the evaluation protocol.