Application of support vector machines in classification of magnetic resonance images

The authors present an application of support vector machines in the classification of magnetic resonance images. Traditional classification methods, such as neural network approaches, have suffered difficulties with generalization, producing models that can overfit the data. The SVM approach is considered a good candidate because of its high generalization performance. At each location in 2D MR images, the SVM classifier is trained through supervised learning to determine which one of the three brain tissues (WM, GM, and CSF) and the background the pixel belongs to. The results are compared with convention methods in terms of the average computational time and the classification error rate. The comparative experiment results demonstrate that the SVM method achieves very good performance with 9.89% classification error rate and 6.18 seconds total time, compared with 11.65% and 173.51 seconds by the FCM and 12.22% and 244.63 seconds by the BP-MLP. The ability of SVM to outperform the BP-MLP and the FCM suggests that SVM is a promising technique in classification of MR images.