Incorporating Adaptive Sparse Graph Convolutional Neural Networks for Segmentation of Organs at Risk in Radiotherapy

Table 2

Comparison between the network applied with the ASGCN and the related approaches.

ROI	DSC of ASGCN	DSC of Swin UNETR	DSC of DCGCN	DSC of SE	DSC of AG	HD of ASGCN	HD of Swin UNETR	HD of DCGCN	HD of SE	HD of AG
Constrictor naris	0.7722	0.7428	0.7738	0.7702	0.7774	1.7650	2.6254	1.7373	1.7634	1.6910
Eyes	0.8951	0.8855	0.8890	0.8968	0.8864	1.1163	1.1728	1.3087	1.1728	1.1584
Lens	0.6495	0.6554	0.6119	0.6454	0.6300	2.4956	1.4060	4.4368	4.3967	5.2654
Optic nerves	0.6418	0.6294	0.6326	0.6220	0.6372	2.3080	2.7100	2.6552	2.8887	2.5706
Temporal lobes	0.8690	0.7939	0.8577	0.8594	0.8579	2.1383	5.8505	3.3170	2.8090	3.6000
Thyroids	0.8089	0.7686	0.7990	0.7965	0.7886	2.0280	2.9508	4.8513	2.3406	2.3329

Higher DSC and lower HD represent superior performance. The bold value denotes better performance.