The proposed work extends previous research performed by this investigator in the field of knowledge-based systems for the analysis of histopathologic material, mainly neoplastic tissue; and mostly focusing on adenocarcinoma of the prostate. The present application concerns the creation of methods for the definition and analysis of novel histopathologic features predominantly derived from nuclear image data, with the objective of defining """"""""prototype identities."""""""" According to the investigator, these are fundamental complexes of histologic features unique to individual tissue samples, or small groups of such samples, that should convey useful predictive value for individual patients. Many of these features are not normally discernable to the eye, even to the experienced observer. They encompass a large number of primary and derived nuclear morphometric measures, including what are referred to as """"""""weak features,"""""""" that is those that have not in the past shown strong correlative utility by standard statistical measures. Derived measures include feature and heterogeneity profiles. These data will form the inputs to a logic network (""""""""inference network"""""""") designed to perform an identification process and ultimately to generate the prototype identities.
