Breast Cancer Detection in Mammogram Images Exploiting GLCM, GA Features and SVM Algorithms

Abstract

This paper presents the novel computing algorithms to maintain the quality of mammogram images for better quality of cancer detection. The advanced algorithms were incorporated with a cancer detection unit to allow an automatic and better accuracy of tumor determination and to better classify the existing normal and abnormal breast tissues. The proposed cancer detection method consists of several steps: The first stage of the Computer Aided Detection is to maintain the images and to show the location of the abnormal tissues. The pre-processing performed on the sampled image utilized the morphology algorithm and the multi threshold segmentation to provide the appropriate tissue classification. The use of the morphology algorithm was optimized to eliminate the presence of the mammogram image label. The textural features analysis was obtained by using Gray Level Coocurance Matrix (GLCM) of four different angles, i.e. 00, 450, 900, and1350, respectively. Genetic Algorithm (GA) was optimized to find the best GLCM features, and then the results were inserted in the Support Vector Machine (SVM) training. SVM with kernel radial basis function was used to classify the patient’s images as normal or abnormal breast. SVM algorithm was very important during the data training and the data testing steps. Interesting results were generated during SVM classification, which include the sensitivity rate of 69%, the precision rate of 100% and the system classification accuracy of 88.2% were taken outside from the training data and 100 % were taken inside the training data.