We collaborate with clinical and computational pathology researchers to develop and evaluate AI methods that extract clinically meaningful information from routine tissue and cytology images. Many important cancer biomarkers (e.g., patterns of immune activity, cellular atypia, or molecular signals associated with recurrence) are difficult to measure using standard pathology workflows or require expensive molecular assays. Our work explores how machine learning and digital pathology can infer these signals directly from commonly collected clinical data, such as histology slides and cytology images, with a current focus on bladder cancer and colon cancer. By combining computational modeling with human-centered design and multicenter clinical studies, we aim to develop tools that help clinicians make more accurate prognostic and treatment decisions while improving the scalability and accessibility of advanced cancer diagnostics.

Dr. Joshua Levy, Dr. Fred Kolling, Dr. Parth Shah