Classification of MRI Images Using Neural Network

Aqsa Korai; Mukhtiar Ali Unar

Aqsa Korai Institute of Information and Communication Technologies, Mehran University of Engineering and Technologies, Jamshoro
Mukhtiar Ali Unar Institute of Information and Communication Technologies, Mehran University of Engineering and Technologies, Jamshoro

Keywords: Brain tumor, MRI, Glioma, Meningioma, Pituitary Tumors, CNN, VGG16, VGG19, Softmax

Abstract

Diagnosing a brain tumor usually begins with Magnetic Resonance Imaging (MRI). However, the complexity associated with the MRI system makes this task non-trivial. Especially, distinguishing between different types of tumors, i.e., Meningioma, Glioma and Pituitary is not easy and is highly subjective. To address this issue, we train a Convolutional Neural Network (CNN) using transfer learning to classifying MRI images into the mentioned tumor types. We used pre-trained models of VGG16 and VGG19 and re-trained them on the publicly available dataset of figshare. A comparison of the performance of these models reveals that VGG16 models achieves a maximum accuracy of 84.59%, whereas the maximum accuracy attained by VGG19 is 86.70%. Our experimental results demonstrate
that the task of classifying tumorous MRI images can be efficiently done with CNN.