Evaluation and Classification of the Brain Tumor MRI using Machine Learning Technique

Abstract

The proposed work implements a Machine-Learning-Technique (MLT) to evaluate and classify the tumor regions into low/high grade based on the analysis carriedout with the brain MRI slices. The MLT implements a sequence of procedures, such as pre-processing, post-processing and classification procedures. The pre-processing enhances the tumor section based on Social Group Optimization (SGO) algorithm assisted Fuzzy-Tsallis thresholding. The robustness of the proposed thresholding is also confirmed by considering the noise corrupted MRI slices.  The post-processing implements the Level-Set Segmentation (LSS) to mine the tumor region. The performance of the LSS is validated with segmentation procedures, like Active-Contour (ACS) and Chan-Vese (CVS) technique. The fundamental data of the tumor section is then extracted using the Gray Level Co-occurrence Matrix (GLCM) and most dominating features are then chosen with a statistical test. Finally, a two-class classifier is implemented using the Support Vector Machine with Radial Basis Function (SVM-RBF) kernel and its performance is then validated with  other classifiers, like the Random-Forest and k-Nearest Neighbor. The outcome of the proposed work confirms that, implemented tool with the SVM-RBF helps to achieve an accuracy of >94% on the benchmark BRATS2015 database.