The diagnosis of hematologic malignancies requires many qualitative ancillary tests to substantiate the impressions made from hematoxylin and eosin stained slides. We propose to use artificial intelligence techniques to expedite and integrate evaluation of information received form histologic morphology, special chemical and enzymatic stains, immunoperoxidase stains, flow cytometry, and molecular diagnostic studies. The goals for this fellowship include: 1) design and implement a database of the important clinical laboratory variables relevant to the diagnosis of hematologic malignancies, 2) develop an expert system based on knowledge-based rules derived from the medical literature and from internal clinical material, and 3) validate the system with retrospective and prospective clinical data. To accomplish these goals and to build this expert system, techniques from two major disciplines in computer science will be applied. First, the relational model of database management will be used to store and retrieve clinical data. Second, artificial intelligence techniques such as predicate calculus, case-based reasoning, neural networks, and fuzzy logic will be used to develop this expert system.
