A Longitudinal Multimodal Risk Prediction Platform for Personalized Breast Cancer Screening

About This Grant

Abstract Breast cancer remains one of the leading causes of cancer-related mortality worldwide, with more than two million new cases diagnosed annually and one in eight U.S. women facing a lifetime risk of the disease. Early detection and precise risk assessment are critical to improving survival, yet current methods are limited in accuracy and accessibility. Current standard of care clinical models based on family history and lifestyle factors fail to identify many at-risk women, and the Breast Imaging Reporting and Data System (BI-RADS) does not fully integrate longitudinal imaging or diverse clinical risk variables. Medara is developing a predictive AI platform powered by the Longitudinal Mammogram Risk (LoMaR) model, an image-based tool designed to predict short-term breast cancer development by leveraging routine mammographic data alongside clinical factors. In this Phase I project, the team will extend the LoMaR framework to a large, diverse, multi-vendor dataset of over 250,000 mammograms linked with biopsy outcomes and clinical risk variables. Aim 1 will improve predictive accuracy through a multimodal longitudinal machine learning model trained on digital breast tomosynthesis (DBT) and clinical risk factors, with performance measured by area under the curve (AUC), falsepositive and false-negative rates, and F1 score. Aim 2 will calibrate LoMaR’s risk scores to clinically interpretable categories aligned with 5-year absolute risk, enabling the generation of personalized screening recommendations. Aim 3 will systematically evaluate the predictive value of longitudinal imaging across DBT, synthetic mammograms, and 2D full-field mammography to guide optimal screening frequency and modality. The anticipated outcome is a functional prototype capable of delivering individualized short-term breast cancer risk assessments with higher precision than existing tools. By integrating multimodal data and longitudinal imaging, this innovation has the potential to reduce unnecessary imaging, enhance early cancer detection, and personalize screening strategies. Ultimately, this project will lay the groundwork for prospective clinical validation and regulatory clearance, supporting the broader goal of improving breast health outcomes through precision screening.