Classification of Brain MRI Using Deep Learning Techniques

Abstract

MR images are popularly used as a tool in diagnosis of brain tumors. It is widely used because of its varieties of angles and clarity of anatomy. Brain tumor is dangerous if it is malignant or secondary tumor. These kinds of tumor can easily spread from one location to another. Expertise and human intervention are needed to detect any kind of abnormalities like tumor, etc., from MR image. So, if we can use an automated brain tumor detection methodology to predict the presence of tumor in brain without human intervention, it will provide an edge in the process of treatment to this disease. Classification plays a vital role in detection of brain tumor. Taking into account the importance of detection of brain tumor, this paper analyzes four architectures of convolutional neural networks (CNN) for classification of brain MR images into tumorous or nontumorous in unsupervised manner. The architectures which are discussed in this paper are ConvNet, Lenet, ResNet, and Densenet.