Brain Tumor Classification Using Efficient Deep Features of MRI Scans and Support Vector Machine

Abstract

Brain cancer is a life-threatening disease and hampers the normal operation of the human body. For correct diagnosis and methodical treatment planning, it is essential to detect the brain tumor in initial stages of development. This study proposes an intelligent diagnostic method for early detection of brain tumor. In the developed method, the deep features of magnetic resonance imaging (MRI) scans are used as the input of support vector machine (SVM). In the first step of the proposed method, Grab cut method is applied for segmenting tumor region, then the segmented images are fed to convolutional neural network (CNN) for deep feature extraction. Following feature extraction module, minimum Redundancy Maximum Relevance (mRMR) algorithm is used to select the most efficient deep features. Finally, the selected deep features are fed into SVM in the classification module. In addition, we applied the black widow optimization algorithm (BWOA) for optimal tuning of hyperparameters of CNN and SVM. The developed method applied on BraTS 2015 datasets and the obtained results showed that the developed method is effective and may be employed in computer-aided diagnosis systems to classify the type of tumor.