LongitudinalMamba: fusing longitudinal changes of mammograms with Mamba for breast cancer diagnosis
Zhengbo Zhou, Dooman Arefan, Margarita L. Zuley, Jules H. Sumkin, Shandong Wu · 2025
In recent years, deep learning has showcased substantial promise in breast cancer diagnosis via mammograms. Although integrating longitudinal changes between consecutive mammograms, which clinicians often rely on for diagnosis, is crucial, many existing methods struggle to efficiently capture these relationships and are computationally intensive. In this study, we introduce an novel method that leverages Mamba method to capture longitudinal information between consecutive mammograms taken at various time intervals. Our method’s efficacy was evaluated using a case-control internal dataset comprising 590 cases. Preliminary results highlight its superiority over models that rely solely on a single ’current’ mammogram exam, as well as those that combine features from both ’current’ and ’prior’ mammograms. By leveraging Mamba for the fusion of ’current’ and ’prior’ mammograms, our model demonstrated enhanced diagnostic performance.