Detection and Grading of Breast Cancer via Spatial Features in Histopathological Images

Abstract

In this study, the detection and grading of breast cancer, which is a very common type of cancer, are studied. In feature extraction phase, feature matrices are created by using local binary pattern histograms and Gabor filters on a dataset with 4812 train data. Afterwards, the classification process are carried out using k-nearest neighbor, decision trees and several ensemble classification methods such as bagging, adaboost, random forests. The accuracies of the classifications are measured with 491 test image which are selected from dataset. At the end of the study, success of feature extraction methods and classifiers are compared and the results were presented in tables.