University of Khartoum

Mutilmodal Mri Brain Tumor Segmentation

Mutilmodal Mri Brain Tumor Segmentation

Show full item record

Title: Mutilmodal Mri Brain Tumor Segmentation
Author: Fadul, Ruba Ali Hamad
Abstract: Magnetic Resonance Imaging (MRI) is widely used in medical imaging to visualize brain tissues. Gliomas brain tumor experiences high heterogeneity in appearance and shape in MRI scans so automated tumor segmentation in brain MRIs is technically challenging. MRI tumor segmentation is important in the preparing of tumor treatment plans. Manual segmentation is time consuming exposed to human errors. A lot of algorithms were made to automate this procedure, yet this field is still considered an open area for research as there is no satisfying accuracies achieved. In this project we conducted a literature review on the previous methods used to segment brain tumors in order to find out an appropriate approach to adapt. After classifing and evaluating previous methods we found that Convolutional Neural Networks achieve considerable enhancements in this field, in the rest of the work patch–wise CNNs were adapted to build our solution. We built upon existing CNN solutions to the Brain Tumor Segmentation (BTS) problem, a deep CNN architecture to perform MRI tumor segmentation and the proposed solution was discussed in details. Our proposed solution achieved a maximum Dice Similarity Coeffecient (DSC) of 0.94 for tumor, however some limitations were discoved and discussed in this thesis. In addition, some possible future works were pointed out.
URI: http://khartoumspace.uofk.edu/123456789/25801
Date: 2017


Files in this item

Files Size Format View

This item appears in the following Collection(s)

Show full item record

Share

Search DSpace


Browse

My Account