Multi-class segmentation of brain tumor using Convolution Neural Network

In this report a fully Convolution Neural Network (CNN) architecture is used to segment multi-modal Brain Tumors from Magnetic Resonance (MR) images. Due to the challenges in manual segmentation, computerized brain tumor segmentation is one of the most important challenges in medical imaging. The fully convolutional structure of the network makes it faster than any network with a dense fully connected layer. The two phase training and entropy sampling of data makes it easier to learn tumor boundaries and overcome the data imbalance problem.

Name thesis:degree_name

Master of Science in Computational Science, Engineering, and Mathematics

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Computational Science, Engineering, and Mathematics

Grantor

The University of Texas at Austin

Year dc:date.issued

2018

Author dc:creator

• Azmat, Muneeza

Advisor dc:contributor.advisor

• Biros, George

• Brain tumor
• Convolution Neural Network
• Segmentation
• Multi-class segmentation
• Multi-modal brain tumors
• Magnetic Resonance images
• Computerized brain tumor segmentation
• Tumor boundaries
• Data imbalance problem

Language dc:language.iso

en

Identifier

doi:10.15781/T27941C1P

OAI identifier oai:identifier

oai:repositories.lib.utexas.edu:2152/65765