Brain Tumor Detection and Classification by SVM Algorithm and Performance Analysis Through CNN Approach

Abstract

A tumor is the uncontrolled growth of tissues inside a part of body. When it grows inside the brain, it is more life threatening. Only proper treatment by detecting tumor in early stage can save human life. In advanced technology, automated computer aided method can be used in early detection of the tumor. In this research, a semi-automatic method comprising an anisotropic filter and a support vector machine (SVM) classifier is used to segment and classify tumors. Firstly, feature extraction algorithm is used to know the condition of brain cells. Next SVM classifier is used to classify the tumor type in a MRI brain image. The proposed SVM algorithm uses twelve features extraction data to analyze the type of tumor. After segmentation and classification through SVM, a machine learning algorithm (e.g. convolutional neural network-CNN) is utilized to verify detection accuracy. A 19-patch based CNN algorithm that include convolution, batch normalization, ReLu, and max-polling layer is proposed in this work. This CNN algorithm takes 70% of data as a train set and 30% of data as a test set. The anisotropic filtering (ADF) based SVM classification model has an accuracy of 97.14%, sensitivity of 98.55% and specificity of 66.67%. The proposed 19-patch based algorithm of CNN gives an accuracy of 94.00%. In compared to state-of-the-art methodologies, the simulation results demonstrate the significance in terms of quality parameters and accuracy.