Neural Networks to Detect and Classify Bladder Cancer

CytoDiagnostics, Inc., and the University of Oklahoma developed partially automated, user-interactive systems for quantitation of biomarkers used in the diagnosis of bladder cancer. The goal of Phase I is to demonstrate the feasibility of further automation with an online neural network system integrated with the quantitative fluorescence image analysis system to classify bladder cancer cells by the pattern of distribution of specific molecules in bladder cancer cells. Researchers at the University of Oklahoma and Tulsa University have collaborated to show the feasibility of first-generation neural network technology to successfully recognize bladder cancer cells from the distribution of DNA within cells. The Phase I project will be a collaboration among CDI and these researchers to initiate development of a commercially feasible system and will (1) develop a representative grey-level image library of cell types of cells labeled for DNA distributions (Hoechst 33258 dye) and a low grade tumor antigen (M344 anti- body), (2) characterize advanced neural networks for discriminating cells off-line, and (3) determine feasibility of developing on-line systems. A Phase II project would develop on-line neural networks integrated with fast image-analysis systems for diagnosis and prognosis of bladder cancer in a commercial diagnostic laboratory.