Fractional sobel filter based brain tumor detection and segmentation using statistical features and SVM

Abstract

In this paper, a scheme for detection and segmentation of brain tumor from T1-weighted and fluid-attenuated inversion recovery (FLAIR) brain images is presented. To prevent the effect of noise and enhance texture of the brain image, fractional Sobel filter is used. Fractional order (α) of fractional Sobel filter provides additional flexibility in improving the segmentation results. Detection of asymmetry between hemispheres is achieved using Bhattacharya coefficients and mutual information. In order to detect the hemisphere containing tumor, histogram asymmetry method is applied. To segment the tumor region from the tumor hemisphere, the statistical features of a defined window are calculated and classified using support vector machine (SVM). Simulations are performed on the images, taken from the BRATS-2013 dataset, and performance parameters such as accuracy, sensitivity, and specificity for different values of α are computed. The simulation results show that the performance of proposed scheme is comparable to the nearest schemes compared.