GAU U-Net for multiple sclerosis segmentation

doi:10.1016/j.aej.2023.04.069

2023-07-01

Roba Gamal *

Hoda Barka

Mayada Hadhoud

DOI：10.1016/j.aej.2023.04.069

原文链接

查看原文

摘要

En 中文

Multiple sclerosis is an auto immune disease which affects the brain and nervous system. A total of 2.8 million people are estimated to live with Multiple sclerosis worldwide (35.9 per 100,000 population). The pooled incidence rate across 75 reporting countries is 2.1 per 100,000 per-sons per year, and the mean age of diagnosis is 32 years. Lesions resulting from the disease can be spotted in the patients MRI scans. In this paper a novel Deep learning architecture GAU-U-net is proposed. The model is inspired from the very famous U-Net architecture used for semantic seg-mentation and widely used in medical image segmentation. The proposed model consists of 3D U-Net after adding a new attention technique inspired by the Global Attention Upsample unit. By using GAU-unet architecture the Dice coefficient increased from 64% to 72% compared to using 3D-Unet.Also, when compared with Unet-attention network the dice coefficient increased from 69% to around 72% with a considerable incline in the number of model parameters in favor of our architecture, which uses 28 M parameters compared to Unet-attention which uses100M parameters.(c) 2023 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Alexandria University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/ licenses/by-nc-nd/4.0/).

Keyword:

Multiple sclerosis

U-Net

MRI segmentation

GAU

Attention

3D U-net