Comparison of features detections and classifiers for breast cancer classification

Breast cancer is one of the most popular diseases which is very harmful especially for women. The method to detect the breast cancer tissue is very important as to aid the radiologist in detecting and diagnosing breast cancer. Computer aided diagnosis (CAD) can be used as one of the systems to help improve the mammographic diagnosis as it can lower false-negative results. A set of 111 abnormal mammographic images from the mini-Mammogram Image Analysis Society (MIAS) database was obtained and used for implementation. Median filter and histogram stretching method are used to pre-process the images. Hybrid Otsu and Mathematical Morphology are used in segmented phases. Then, the features are detected by using both Oriented FAST and rotated BRIEF (ORB) and Binary Robust Invariant Scalable Keypoints (BRISK) as feature detection and feature extraction. A comparison of ORB and BRISK features performance are made as different classifiers, which are Fine KNN and Support Vector Machine (SVM) are also used to classify the breast tissue. The highest accuracy for both ORB and BRISK features is achieved by using SVM classifiers which is 94.5% and 99.4% for testing dataset. However, Fine KNN classifiers are faster in terms of the prediction speed despite having lower accuracy than SVM.