Machine vision-based histometry of premalignant and malignant prostatic lesions.

The implementation of knowledge-guided control of the processing and segmentation of histopathologic images of prostatic lesions has made automated analysis and interpretation possible. To establish correspondence between histopathologic concepts, terms and diagnostic criteria, and computed histometric entities, "interpretive transforms" are introduced. Scene segmentation is controlled by an expert system following a model-based reasoning process. The expert system is structured as an associative network with frames at each node, which controls a knowledge file and a large library of image processing algorithms.