TY - GEN
T1 - BI-GCN: Boundary-Aware Input-Dependent Graph Convolution Network for biomedical image segmentation
AU - Meng, Yanda
AU - Zhang, Hongrun
AU - Gao, Dongxu
AU - Zhao, Yitian
AU - Yang, Xiaoyun
AU - Qian, Xuesheng
AU - Huang, Xiaowei
AU - Zheng, Yalin
N1 - No ISSN
DBLP License: DBLP's bibliographic metadata records provided through http://dblp.org/ are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.
PY - 2021/10/15
Y1 - 2021/10/15
N2 - Segmentation is an essential operation of image processing. The convolution operation suffers from a limited receptive field, while global modelling is fundamental to segmentation tasks. In this paper, we apply graph convolution into the segmentation task and propose an improved \textit{Laplacian}. Different from existing methods, our \textit{Laplacian} is data-dependent, and we introduce two attention diagonal matrices to learn a better vertex relationship. In addition, it takes advantage of both region and boundary information when performing graph-based information propagation. Specifically, we model and reason about the boundary-aware region-wise correlations of different classes through learning graph representations, which is capable of manipulating long range semantic reasoning across various regions with the spatial enhancement along the object's boundary. Our model is well-suited to obtain global semantic region information while also accommodates local spatial boundary characteristics simultaneously. Experiments on two types of challenging datasets demonstrate that our method outperforms the state-of-the-art approaches on the segmentation of polyps in colonoscopy images and of the optic disc and optic cup in colour fundus images.
AB - Segmentation is an essential operation of image processing. The convolution operation suffers from a limited receptive field, while global modelling is fundamental to segmentation tasks. In this paper, we apply graph convolution into the segmentation task and propose an improved \textit{Laplacian}. Different from existing methods, our \textit{Laplacian} is data-dependent, and we introduce two attention diagonal matrices to learn a better vertex relationship. In addition, it takes advantage of both region and boundary information when performing graph-based information propagation. Specifically, we model and reason about the boundary-aware region-wise correlations of different classes through learning graph representations, which is capable of manipulating long range semantic reasoning across various regions with the spatial enhancement along the object's boundary. Our model is well-suited to obtain global semantic region information while also accommodates local spatial boundary characteristics simultaneously. Experiments on two types of challenging datasets demonstrate that our method outperforms the state-of-the-art approaches on the segmentation of polyps in colonoscopy images and of the optic disc and optic cup in colour fundus images.
UR - https://britishmachinevisionassociation.github.io/bmvc
UR - https://arxiv.org/abs/2110.14775
M3 - Conference contribution
BT - Proceedings of BMVC 2021
PB - British Machine Vision Association
T2 - 32nd British Machine Vision Conference
Y2 - 22 November 2021 through 25 November 2021
ER -