Digital Knowledge and Diagnostic Information

The long term objective of this research was the development of objective, digitally defined procedures for histopathologic assessment. The development of procedures based on digital knowledge had as its first aim the design of a machine vision system with image understanding capability (ie, capable of autonomous processing and analyses of histopathologic imagery). Next, histometric and karyometric diagnostic information extraction led to highly specific characterization of nuclei and lesions. Based on such detailed characterizations, we were able to derive progression curves for prostatic, colonic, breast epithelial, and esophageal lesions. The specific signatures of nuclei and lesions revealed s~tbstantial diversity among lesions of the same visual-diagnostic grade; profiles of deviation of nuclei from a normal standard were derived to provide a novel, additional level of diagnostically discriminating features. Knowledge guided machine vision opens the way to an extremely specific characterization of nuclei and lesions, which may allow better prediction of biological behavior and, thus, more accurate individual patient targeted prognosis. (The J Histoteclzizol 23: 183, 2000)