Polyp segmentation algorithm combining multi-scale attention and multi-layer loss

Junjiang Liu; Changming Zhu

doi:10.1117/12.2652907

10 November 2022 Polyp segmentation algorithm combining multi-scale attention and multi-layer loss

Junjiang Liu, Changming Zhu

Proceedings Volume 12331, International Conference on Mechanisms and Robotics (ICMAR 2022); 123314W (2022) https://doi.org/10.1117/12.2652907
Event: International Conference on Mechanisms and Robotics (ICMAR 2022), 2022, Zhuhai, China

Abstract

Colon cancer is one of the most common malignant tumors in human digestive tract, with high fatality and morbidity. Therefore, early screening of intestinal polyps to judge the probability of colon cancer has very important medical significance for the prevention and treatment of modern colon cancer. In this paper, a polyp segmentation algorithm combining multi-scale attention and multi-layer loss is proposed for the segmentation of intestinal polyp lesions during polyp screening. This algorithm is based on the improved U-NET network. On this basis, it combines DenseNet and adds multi-scale input based on multi-linear interpolation, coordinate attention mechanism and Dice loss to the network to jointly improve the joint segmentation performance of the model. The experimental results of the proposed algorithm on an open data set are better than those of U-NET and other algorithms, and the experimental results show that the proposed algorithm can effectively segment the lesion region of intestinal polyps with good segmentation performance.

Citation Download Citation

Junjiang Liu and Changming Zhu "Polyp segmentation algorithm combining multi-scale attention and multi-layer loss", Proc. SPIE 12331, International Conference on Mechanisms and Robotics (ICMAR 2022), 123314W (10 November 2022); https://doi.org/10.1117/12.2652907

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