An Mr Brain Images Classifier via Principal Component Analysis and Kernel Support Vector Machine

摘要: Automated and accurate classification of MR brain images is extremely important for medical analysis interpretation. Over the last decade numerous methods have already been proposed. In this paper, we presented a novel method to classify given image as normal or abnormal. The proposed first employed wavelet transform extract features from images, followed by applying principle component (PCA) reduce dimensions features. reduced were submitted kernel support vector machine (KSVM). strategy Kfold stratified cross validation was used enhance generalization KSVM. We chose seven common diseases (glioma, meningioma, Alzheimer’s disease, disease plus visual agnosia, Pick’s sarcoma, Huntington’s disease) abnormal brains, collected 160 (20 140 abnormal) Harvard Medical School website. performed our with four different kernels, found that GRB achieves highest accuracy 99.38%. LIN, HPOL, IPOL 95%, 96.88%, 98.12%, respectively. also compared those literatures in decade, results showed DWT+PCA+KSVM still achieved best results. averaged processing time 256× 256 size on laptop P4 IBM 3GHz processor 2 GB RAM 0.0448 s. From experimental data, effective rapid. It could be applied field can assist doctors diagnose where patient certain degrees. Received 14 June 2012, Accepted 23 July Scheduled 19 August 2012 * Corresponding author: Yudong Zhang (zhangyudongnuaa@gmail.com).