Generic placeholder image

International Journal of Sensors, Wireless Communications and Control


ISSN (Print): 2210-3279
ISSN (Online): 2210-3287

Research Article

Brain Segmentation Using Deep Neural Networks

Author(s): Vandana Mohindru*, Ashutosh Sharma, Apurv Mathur and Anuj Kumar Gupta

Volume 11, Issue 1, 2021

Published on: 28 July, 2020

Page: [81 - 88] Pages: 8

DOI: 10.2174/2210327910999200728145536

Price: $65


Background: The determination of the tumor extent is a major challenging task in brain tumor planning and quantitative evaluation. Magnetic Resonance Imaging (MRI) is one of the non-intellectual technique that has emerged as a front-line diagnostic tool for a brain tumor with non-ionizing radiation.

Objectives: In Brain tumors, Gliomas is the very basic tumor of the brain; they might be less aggressive or more aggressive in a patient with a life expectancy of not more than 2 years. Manual segmentation is time-consuming, therefore we use a deep convolutional neural network to increase the performance, which is highly dependent on the operator's experience.

Methods: This paper proposed a fully automatic segmentation of brain tumors using deep convolutional neural networks. Further, it uses high-grade gliomas brain images from BRATS 2016 database. The suggested work achieves brain tumor segmentation using tensor flow, in which, the anaconda frameworks are used to execute high-level mathematical functions.

Results: Hence, the research work segments brain tumors into four classes like edema, nonenhancing tumor, enhancing tumor and necrotic tumor. Brain tumor segmentation needs to separate healthy tissues from tumor regions, such as advancing tumor, necrotic core, and surrounding edema. We have presented a process to segment 3D MRI image of a brain tumor including their separate sub-areas. We have applied an SVM based classification. Categorization is complete using a soft-margin SVM classifier.

Conclusion: Deep convolutional neural networks have been used to present the brain tumor segmentation. Outcomes of the BRATS 2016 online judgment method assure us to increase the performance, accuracy, and speed with our best model. The fuzzy c-mean algorithm provides better accuracy and trains the SVM based classifier. We can achieve the finest performance and accuracy by using the novel two-pathway architecture, i.e.., encoder and decoder, as well as the modeling local label that depends on stacking two CNN’s.

Keywords: Brain segmentation, BRATS dataset, fuzzy c-means, deep neural network, magnetic resonance images, SVM.

Graphical Abstract

Rights & Permissions Print Export Cite as
© 2023 Bentham Science Publishers | Privacy Policy