SVM-kNN and Pixel Gradient Based Segmentation Technique for Brain MRI Tumor Images Classification

Abstract: The research paper proposes a novel and robust segmentation technique for the classification of Brain MRI Tumor Images. It is a completely automatic technique for the segmentation of tumor in a tumor pretentious brain MRI image with its classification from a normal image. In this technique, method is able to detect brain tumor utilizing the pixel intensities differences among the tumor and its neighbouring tissues. Here, it also uses the associated tumor�s pixels labelling which can efficiently segments the brain tumor from the image. After that, SVM-kNN is exploited for the classification of brain MRI image from the normal one. For the feature vector�s computation, GLCM technique is utilized. From the results, it can be observed that with only less number of feature vectors, in terms of speed and computational power used, the method is much better. As accuracy of nearly 98% is computed using the proposed method which is very good in comparison to the other methods used.

Keywords: GLCM, MRI, Solidity and Thresholding, SVM-kNN.