Glioma Brain Tumors Diagnosis and Classification in MR Images based on Convolutional Neural Networks

Brain tumor analysis is a critical field in medical image processing. Glioma is one of the threatening brain tumors originating from glial cells and is divided into two grades according to the World Health Organization (WHO). In this paper, a novel method based on Convolutional Neural Networks (CNN) is presented to diagnose and classify Glioma tumors in Magnetic Resonance Imaging (MRI) images into three classes: Normal Brain, High-Grade Glioma and Low-Grade Glioma. The proposed method includes 2 parts: preprocessing unit and network. Preprocessing unit extracts brain from skull and the obtained image is fed into a CNN network to be classified. The network extracts primary features from images and creates feature maps. Then the second part of the network extracts secondary features from the feature maps and finally classifies them. The datasets used in this paper are IXI dataset as normal brain images and BRATS2017 dataset as Glioma tumor images. This method classifies the MRI images into three categories, performed with a desirable accuracy of 99.18%.