Title: Deep learning techniques for classification of brain MRI

Authors: Imayanmosha Wahlang; Pallabi Sharma; Sugata Sanyal; Goutam Saha; Arnab Kumar Maji

Addresses: Department of Information Technology, North-Eastern Hill University, Shillong, Meghalaya, India ' Department of Computer Science and Engineering, National Institute of Technology, Shillong, Meghalaya, India ' School of Technology and Computer Science, Tata Institute of Fundamental Research, Mumbai-400005, Maharashtra, India ' Department of Information Technology, North-Eastern Hill University, Shillong, Meghalaya, India ' Department of Information Technology, North-Eastern Hill University, Shillong, Meghalaya, India

Abstract: Several brain diseases are becoming a threat to the livelihood of people. One such problem is the presence of a brain tumour. A brain tumour can be benign or malignant. It is dangerous if it is a malignant or secondary tumour (metastasis). Therefore, there is a need to detect the presence of tumours at the earliest stage as possible. Using an automated method for brain tumour detection can be a solution to medical expertise as a biopsy can be excluded if early detection could be possible. Classification helps in the prediction of the type of image and type of tumour. In this paper, three stages are involved. In the first stage, the classification of brain MR images into normal (tumour) or abnormal (non-tumour) images using ConvNet, LeNet, ResNet, and DenseNet has been analysed. In the second stage, architectures like LeNet and AlexNet are used in the prediction of the type of tumour namely metastasis, glioma, and meningioma. And lastly, using U-Net and AlexNet, classification into high grade glioma and low grade glioma was done.

Keywords: convolutional neural network; CNN; DenseNet; ResNet; AlexNet; U-Net.

DOI: 10.1504/IJISTA.2020.112441

International Journal of Intelligent Systems Technologies and Applications, 2020 Vol.19 No.6, pp.571 - 588

Received: 30 Apr 2019
Accepted: 28 May 2020

Published online: 15 Jan 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article