Intelligent classification technique of human brain MRI with efficient wavelet based feature extraction using local binary pattern

An intelligent classification technique for MR brain images are extremely important for medical analysis and treatment selection. Manual interpretation of these images by physicians may lead to wrong diagnosis when a large number of MRIs are analyzed. In this paper an automated decision support system for classification is proposed. It consists of computing the uniform LBP with mapping. Haar wavelet is used to extract the coefficients from the image which are reduced by PCA. These features are given as input to the SVM classifier with three different types of kernel. The proposed system is efficient for the classification of brain images into normal or abnormal with a high accuracy of 91.25 for Linear kernal and 86.25 for polynomial kernel.