CLASSIFICATION OF MEDICAL IMAGES BASED ON THE SPECTRA OF LOCAL WINDOWS
摘要
A method for automated classification of medical images using neural networks is presented. The method includes an image segmentation stage at which areas of interest are selected and a stage at which segments are classified into defined classes of pathological formations. Algorithmic and software support has been developed for medical image classification; experimental testing using control sets for the differential diagnosis of segments of the Oncology and Pancreatitis classes on endoscopic images of the pancreas showed that the diagnostic sensitivity for these classes was at least 0.8 and diagnostic specificity was least 0.85.