Computer aided breast cancer diagnosis system with fuzzy multiple-parameter support vector machine

Abstract

Computer Aided Diagnosis (CAD) system has been proven that it can be utilized as the secondary option for physicians for early breast cancer detection. A typical CAD system consists of several phases like image segmentation, feature extraction and selection, classification. Among those phases, the classification phase is one of the important phases that directly affect the performance of the entire system. Therefore the main issue is to enhance the classification phase to construct better decision-making procedure comparing to conventional classification phase by assigning enhanced logic. In this paper, we propose a Fuzzy Multiple-parameter Support Vector Machine (SVM), which will be used in the CAD system. The proposed method uses fuzzy membership to tune up each training data points by assigning proper weight, corresponding to its feature, and adopts multiple parameters as a classifier for SVM, which further improves the machine-learning process to a more robust level. The experimental result shows that the proposed method is far more superior to the existing SVM in terms of performance, sensitivity and accuracy. Additionally, the result suggests for more sophisticated and complex approach to the current classification for CAD system.