The goal was to translate into commercial practice research by a team at Carnegie Mellon University (CMU) and the University of Southern California (USC) in automating the subtyping of follicular cancer of the lymph node by computerized microscopy. This solves the problem that visual subtyping (observing follicular cells through the microscope and scoring them as small, mixed, or large) is so inaccurate that 30% of the time the treatment plan recommended for the patient is wrong. With about 35,000 individuals developing such cancers annually in the USA each year and this number increasing yearly due to AIDS, the implication was that thousands of patients were receiving the wrong treatment. In the 1980s CMU demonstrated that a version of the $150,000 Coulter Biomedical diff3/50 robot image-analyzing microscope controlled by a minicomputer could subtype with an accuracy of about 2%. In the Phase I proposal the PI suggested a demonstration where the CMU methodology would be moved to a low-cost, PC-based workstation and tested using a new multimode, multicolor camera jointly developed by the Pathology Imaging Corp. and Kensal Consulting (KSC). This was done, 1200 images were digitized and analyzed, and the new workstation divided the small-cell-large-cell continuum into five regions (small, small/mixed, mixed, etc.) with a repeatability of 95.5%. However, the computer time required per specimen was one hour - too slow for convenient clinical use. This Phase II proposal is for funding to accelerate this process 1000 fold by creating and commercializing a for use by the DoD on the SDI program. It is also proposed to demonstrate that the same workstation will analyze cells in immuno-stained breast tumor tissue at almost flow cytometry rates (2000 cells per second) for use in determining the proliferative index.
