MULTI-FEATURES CLASSIFICATION OF PROSTATE CARCINOMA OBSERVED IN HISTOLOGICAL SECTIONS

摘要

The present invention relates to a method for multi-characteristic classification of prostate cancer species in histological sections, the method comprising the steps of staining a histological section of a collected prostate biopsy sample image to generate an RGB input image, with respect to the stained RGB input image A color characteristic extraction step of extracting color characteristics, a texture characteristic extraction step of extracting texture characteristics with respect to the dyed RGB input image, a step of identifying and selecting important features for the extracted color and texture characteristics, and important It includes a classification step of classifying the prostate cancer type using a multilayer perceptron (MLP) neural network classification algorithm based on the combination of color and texture feature values extracted in the feature identification and selection step.