Improved TransUNet retinal vessel segmentation method

Chengxia Zang; Hui Peng

doi:10.1117/12.2656622

20 October 2022 Improved TransUNet retinal vessel segmentation method

Chengxia Zang, Hui Peng

Proceedings Volume 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022); 124511S (2022) https://doi.org/10.1117/12.2656622
Event: 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 2022, Chongqing, China

Abstract

Observing the fundus vessels in a noninvasive manner can play an auxiliary role in the clinical diagnosis of a variety of diseases. It is inevitable that misdiagnosis will occur by means of manual analysis. Aiming at the problems that the accuracy of computer-aided segmentation of fundus vessels is not high and the use of convolution layer sampling will cause the loss of global context information, an improved fundus vessel segmentation network based on TransUNet is proposed. Firstly, the transformer layer of TransUNet can effectively obtain the connection of global context information. Secondly, the convolution unit sampled under TransUNet is replaced by residual space attention block (RSAB) to improve the attention to the vascular region and prevent the information from disappearing and over fitting in the process of passing through the network. Finally, mini-UNet and TransUNet are cascaded to make the model correct the segmentation results and improve the accuracy of vascular segmentation. The experimental results show that the improved algorithm improves the accuracy of drive dataset by 0.23% compared with UNet network. Each evaluation index is better than most of the compared segmentation models.

Citation Download Citation

Chengxia Zang and Hui Peng "Improved TransUNet retinal vessel segmentation method", Proc. SPIE 12451, 5th International Conference on Computer Information Science and Application Technology (CISAT 2022), 124511S (20 October 2022); https://doi.org/10.1117/12.2656622

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
5 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Image segmentation

Transformers

Convolution

Image processing

Blood vessels

Data modeling

Performance modeling

Show All Keywords

Keywords/Phrases

Search In:

Publication Years