Strategic Initiative Fund (SIF)

AI-Powered Optical Biopsy for Real-Time In Vivo Pathology

The AI-Powered Optical Biopsy for Real-Time In Vivo Pathology initiative would enable clinicians to evaluate tissue microscopically during routine endoscopy rather than relying solely on physical biopsies and off-site pathology. U-M researchers are developing fiber-coupled dual-axes confocal endomicroscopes that produce high-resolution images resembling conventional histology. Initially focused on colorectal cancer and inflammatory bowel disease–associated dysplasia, the technology aims to detect disease earlier, improve point-of-care decisions, and reduce unnecessary biopsies, procedural delays, costs, and inequities in access to diagnostic care.

The project has three scientific and educational aims: create clinically deployable endomicroscopes, develop AI algorithms that interpret optical biopsies in real time, and train clinicians, engineers, and scientists to use the technology responsibly. The team will move from engineering and early clinical studies to broader clinical deployment, education, and commercialization. Beginning with at least two systems and more than 500 patients, the program will progressively expand clinical deployment, patient enrollment, and institutional capacity toward high-volume routine use.

Images collected during endoscopy will be linked to pathology and clinical data in the Michigan Optical Biopsy Atlas, a governed U-M resource supporting algorithm development, validation, regulatory submissions, and future research. Deep-learning models will identify tissue features across normal, precancerous, cancerous, and inflammatory conditions while being tested for accuracy, robustness, generalizability, bias, and interpretability. The long-term goal is a permanent, institution-wide clinical and data platform embedded within Michigan Medicine and expandable to other organs, hospitals, and partner institutions.