Yudong Zhang1, 2, Chuanmiao Xie, 13, Bradley S. Peterson1, 2, Zhengchao Dong1, 2
1Columbia University, New York, NY, United States; 2New York State of Psychiatric Institute, New York, NY, United States; 3Department of Medical Imaging & Interventional Radiology, Sun Yat-Sen University, Guangzhou, Guangdong, China
We proposed a novel hybrid system to classify an MR brain image as either normal or abnormal. The method employed digital wavelet transform to extract features and used principal component analysis to reduce the dimensionality of the feature space. Afterwards, we constructed a kernel support vector machine with Radial Basis Function kernel, using particle swarm optimization to optimize the parameters in the training function. We tested the method with a dataset of 90 brain images consisting of 17 diseases. A 5-fold cross validation showed that our method achieved 97.78% classification accuracy, higher than 86.22% by BP-NN and 91.33% by RBF-NN.