 In this paper, we present a novel computer-aided diagnosis, CAD, system for distinguishing abnormal brains from normal brains in magnetic resonance imaging, MRI. First, discrete wavelet packet transform, DWPT, was applied to extract wavelet packet coefficients from MRI scans. Then, Shannon entropy, SE, and Salis entropy, TE, were utilized to calculate entropy features from the DWPT coefficients. Finally, two types of support vector machines, SVMs, generalized eigenvalue proximate SVM, JEPSOM, and JEPSOM with radial basis function, RBF, kernel, were employed as classifiers. The proposed method was evaluated using three benchmark data sets, data set 66, data set 160, and data set 255. The results show that the proposed DWPT plus TE plus JEPSOM plus RBF method outperformed all other methods in terms of classification accuracy. Furthermore, the offline learning time was 8.443 seconds, and the online prediction time was 0.1059 seconds. This article was authored by Yudong Zhang, Xingqiao Dong, Shuihua Wang, and others. We are article.tv, links in the description below.