Categorization of Brain Tumors using SVM with Hybridized Local-Global Features

Automating the process of tumor detection and classification from brain MRI images is an exciting but difficult due to the difference in the appearance of the tumor tissues from person to person and in many cases, tumors look almost same as the normal brain tissues. An effective classification can be expected by employing better feature extraction methods that can describe the tumor in totality. The paper does a selection and the feature extraction process engaging the statistical features of both segmented and unsegmented images that are combined to form a hybrid feature set. These hybrid features is employed to build SVM classifier model. The SVM is modelled as a non-linear classifier with different kernel functions like Iinear, quadratic and RBF for categorization of brain tumors in to one of the two classes namely benign and malignant.