An MR brain images classification technique via the Gaussian radial basis kernel and SVM

Abstract

Computer-aided diagnosis (CAD) and artificial intelligence (AI) are hot topics in the field of clinical imaging and neuro-imaging. Recently numerous methods were proposed. In this research work, a new 3D magnetic resonance head images (MRI) classifier based on KPCA and SVM is presented. The proposed algorithm, called the support vector machine with kernel principal component analysis (SVM-KPCA), aims to classify an MR brain image as normal or pathological. The system first employed the Discrete Wavelet Transform (DWT) to extract features from the images. After feature vector normalization the kernel principal component analysis (KPCA) is applied to reduce the dimensionality of features. The reduced features were then submitted to a support vector machine (SVM). The strategy of k-fold cross-validation was used to enhance generalization of the proposed algorithm. Seven common brain diseases have been used (Alzheimer's disease, Alzheimer's disease plus visual agnosia, glioma, meningioma, Huntington's disease, sarcoma and Pick's disease) as pathological brains, and MR brain images have been collected from `Harvard Medical School' website and `Open Access Series of Imaging Studies (OASIS)' website, to validate the proposed algorithm. Simulation results were compared with the existing algorithms and it was observed that the proposed work outperforms other algorithms. Working on the same dataset in term of accuracy, sensitivity and specificity.