Performance Improvement of Breast Cancer Diagnosis based on Mammogram Images using Feature Extraction and Classification Methods
Elvira Sukma Wahyuni, Retno Paras Rasmi, Suatmi Murnani · 2021
Breast cancer is the most common cause of death in the elderly women category. In 2018, there were almost 0.63 million cases of death found in 2.09 million new cases of breast cancer. Early detection is a key to lower the rate of death caused by breast cancer. One of the most typical methods for early detection is utilizing digital image processing on mammogram images. The detection process has three main steps, namely preprocessing, feature extraction, and classification. Feature extraction plays a significant role in providing an accurate detection system. In this study, we evaluated three different methods of feature extraction. They are feature extraction based on texture, gray level co-occurrence matrix (GLCM), and morphology. In the classification step, we compared the performance of K-Nearest Neighbor, Naive Bayes, and Support Vector Machine methods. We used a mammogram dataset from the Mammographic Image Analysis Society (MIAS) consisting of 100 images labeled as normal and 51 images specified as abnormal. The experimental results show that combining all extracted features and using Naive Bayes classifier obtained the highest accuracy of 98.68%. The results suggest that early detection of breast cancer can be performed accurately using appropriate detection methods.