University of Khartoum

Mutilmodal Mri Brain Tumor Segmentation University of Khartoum

Mutilmodal Mri Brain Tumor Segmentation University of Khartoum

Show full item record

Title: Mutilmodal Mri Brain Tumor Segmentation University of Khartoum
Author: Mohamed, Mohamed Babikir Ali
Abstract: Brain tumor is one of the deadliest diseases human had ever faced, the segmentation of tumors is an important step in the evaluation of tumor, preparing the treatment plan and estimating the patient survial period, manual brain tumor segmentation is a time consuming task and is exposed to human errors. In the past few years many solutions were proposed to automate brain tumor segmentation,in this work many of these solutions were reviewed and evaluated to determine the most promising approaches to obtain accurate results and acceptable segmentation time. In this work we used convolutional neural network to preform the segmentation task, we tried different CNN architectures to preform segmentation and then focused on optimizing patch-wise classifier CNN. The results obtained during this project are discussed to show the effect of some design decision taken, highlight the advantages and limitations of the final model and to point out possible further improvements to it. The final CNN architecture provided a maximum of 0.94, 0.92 Dice similarity coefficient in the whole and core regions respectively, supporting the notion of great CNN potential in this field, the possible future work for improving this architecture and building new solutions to solve this problem were also discussed.
URI: http://khartoumspace.uofk.edu/123456789/25775
Date: 2017-10


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