Classification of Histopathological Images by Spatial Feature Extraction and Morphological Methods

Abstract

The high accuracy of the computerized analysis of histopathological images is very important in the detection of cancerous cells. Thanks to the images with high accuracy, early diagnosis will be made with the detection of cancerous cells. Four different types (benign, normal, in situ carcinoma, invasive carcinoma) classification performances will be analyzed by applying various methods to cancer cells. At the beginning of the studies, the BACH data set was obtained, then the desired and usable parts were tried to be extracted with image processing methods. After obtaining data and images of different sizes, their features were extracted with different algorithms (HOG, GLCM, EMP, SIFT, SURF, LBP), and then the accuracy of classifications was examined with RF, KNN, SVM machine learning algorithms and transfer learning algorithm ResNet.