Hsian-Min Chen1, Jyh-Wen Chai2, San-Kan Lee2, Clayton Chi-Chang Chen2, Ying-Cheng Lin3, Yen-Chieh Ouyang3, Chein-I Chang4, Wu-Chung Shen1
1Department of Radiology, China Medical University Hospital, Taichung, Taiwan; 2Department of Radiology, Taichung Veterans General Hospital, Taichung, Taiwan; 3Department of Electrical Engineering, National Chung Hsing University, Taichung, Taiwan; 4Remote Sensing Signal and Image Processing Laboratory, Department of Computer Science and E.E., University of Maryland, Baltimore County, USA
Support vector machine (SVM) has been widely used as a powerful tool for classification problem arising from various fields and shown that the parameters are critical in the performance of SVM [1]. However, the same parameters are not suitable for all classification problems. In this paper, numerical results show that the performance of SVM with optimal parameters is significant difference to empirical parameters. In addition, we recommend independent component analysis (ICA) method as the pre-processing step to get the robust performance of SVM classification problems in brain MRI.