My reserach centers around integrating longitudinal and multimodal information to improve the performance and generalizability of statistical models in predicting lung cancer diagnosis and risk. We test this hypothesis in two parts. In the first part, we develop approaches to learn lung cancer diagnosis from longitudinal data. In the second part, we seek to combine chest imaging with electronic health records (EHR) longitudinally. We align with both the innovation and application goals of VALIANT through developing multimodal learning methods and practical application to lung cancer management.