PROXIMA: PRecision Oncology through eXploitation of Imaging-guided Medical AI

Prostate cancer is a highly heterogeneous disease, with variations at both the cellular (microscale) and whole-body (macroscale) levels, posing challenges for accurate diagnosis and treatment planning. While histopathology and spatial biology provide insights into tumor microstructure, PET/CT imaging captures systemic tumor behavior. However, current approaches lack integration across these scales, limiting our understanding of how cellular features influence whole-body imaging signals [1]. To move beyond static, single-source imaging snapshots toward dynamic, predictive models that integrate diverse biological and clinical data, this project proposes an AI-driven, physics- informed models to link microscale and macroscale imaging, enabling a more comprehensive characterization of prostate cancer. By integrating multi-modal data and generative AI, we aim to improve tumor profiling, therapy response prediction, and clinical decision-making. This integration will also allow the generation of missing data, reducing the need for invasive overtesting and enabling clinical support even when acquiring information is constrained by time or resources.