Multikernel Support Vector Machine with GLCM For Truthful Brain Signal Classification of Magnetic Resonance Imaging

Abstract

This paper works on the novel hierarchical transformation technique (HTT)on the root of gray level co-occurrence matrix (GLCM) texture features which has to be extracted and sequentially processed with multi-kernel support vector machine (SVM) classifier to categorize MRI brain image clinical states. This processing technique possess 3 different phases. First, pre-processing steps are applied to enhance the image quality by applying hierarchical transformation technique. Then, texture features are extracted with the help of GLCM. Finally, multi-kernel support vector machine algorithm enables the classification. The novel proposed HTT methodology makes the incorporation of optimal selection of mask with disk shaped bottom and top hat morphological processing and few mathematical processes for the cumulative enhancement and the pre-processing of the image. The computation of GLCM is being made for extracting the probabilistic texture features. These features are measurement of contrast, energy, correlation, entropy and homogeneity. The extracted GLCM texture features are based on co-occurrence and then applied with SVM classification which can categorize the acquired MRI brain image into two clinical states as abnormal and normal. Additionally, the enumeration of comparison is also made with traditional feature extraction technique using the GLCM textures.