Certain investigations on classification and analysis of breast image for implementation of computer aided decision support system

Abstract

Mammography has been proven to be the most powerful and reliable tool for the early detection of breast cancer in women who have no symptoms or signs of the disease. The mammogram confirms whether the changes in breast region are due to the presence of benign (non-cancerous) and no treatment is needed, or malignant (breast cancer}.In this research work, the cancer regions are detected and segmented using soft computing technique as the classification method. The proposed methodology for cancer region segmentation consists of the following stages: preprocessing, classifications, feature extraction, segmentation and classifications. The preprocessing stage consists of noise filtering and enhancement. The features are extracted from the enhanced mammogram image and then these features are classified using Extreme Learning Machines (ELM) classification algorithm, which classifies the source mammogram image into either normal or abnormal. The cancer regions are segmented in abnormal image using morphological operations and the finally the performance metrics are used to evaluate the performance of the proposed breast cancer detection methodology. This research work also describes the features to be extracted from the mammogram image with feature optimization techniques. The impact of extracted and optimized set of features on the classification results are also analyzed for improving the classification rate. The optimized set of features from the mammogram image are trained and classified by differenclassification approaches. The classification technique classifies the mammogram image into either normal image or malignant image based on the extracted and optimized set of features. The performance of the proposed breast cancer region detection system is analyzed with respect to various performance evaluation metrics as sensitivity, specificity and accuracy. The simulation results of this proposed method is compared with other conventional methods with respect to various performance evaluation parameter newline