VGU RESEARCH REPOSITORY
Please use this identifier to cite or link to this item:
https://epub.vgu.edu.vn/handle/dlibvgu/991
Title: | Liver segmentation using fully convolution neural networks | Authors: | Ho Quoc Dat | Issue Date: | 2020 | Abstract: | Medical segmentation is a promising field in semantic segmentation in the current years due to the explosive growth of computational power and development of many deep neural networks. Liver segmentation, being a branch of medical segmentation, has its place for being critical in helping with early diagnosis of liver-related diseases. Fully automated liver segmentation is also vital because of the limited manpower of highly trained professionals that usually perform the task of liver segmentation. In addition, segmenting thousands of slices done by professionals is a time-consuming and error-prone process which is desperately in need for a better, less painstaking method. Our project focuses on this aspect to develop a fully automated liver segmentation tool that can execute the liver segmentation with competitive accuracy and low error by using a hybrid of residual network with fully convolutional network - ResUNet as the baseline model. ResUNet was trained using a combination of MICCAI LiTS challenge dataset and 3DIRCAD dataset. Despites having limited computational power, I was able to achieve approximately 91% and 87% dice coefficient score on the liver segmentation of 3DIRCAD dataset and MICCAI LiTS dataset, respectively. |
URI(1): | http://epub.vgu.edu.vn/handle/dlibvgu/991 |
Appears in Collections: | Computer Science (CS) |
Files in This Item:
File | Description | Size | Format | Existing users please Login |
---|---|---|---|---|
Liver segmentation using fully convolution neural networks.pdf | 9.08 MB | Adobe PDF |
Page view(s)
79
checked on Apr 20, 2024
Download(s)
35
checked on Apr 20, 2024
Google ScholarTM
Check
Items are protected by © Copyright of Vietnamese - German University Library