Advancing Biomarker Discovery in Prostate Cancer Through Computer Vision-Driven Profiling of Tumor and Microenvironmental Heterogeneity
Prostate cancer is the most common non-skin cancer in men in the United States and the second leading cause of cancer death. A major challenge in treating prostate cancer is distinguishing slow-growing tumors that may never cause harm from aggressive tumors that require intensive treatments. Current tools, including prostate-specific antigen (PSA) levels, Gleason score, and tumor stage, provide important information but cannot always reliably make this distinction. This can lead to overtreatment for some patients and undertreatment for others. Our research uses artificial intelligence (AI) to uncover hidden patterns in routinely collected medical images, including digitized pathology slides, MRI, and prostate-specific membrane antigen-positron emission tomography (PSMA-PET) scans. By analyzing images from thousands of patients, we aim to identify features linked to tumor aggressiveness, the immune environment and patient outcomes. Our goal is to develop accessible, interpretable AI tools that can be integrated into routine clinical workflows, helping personalize treatment decisions, reduce unnecessary side effects and improve outcomes for patients with prostate cancer.
