TY - GEN
T1 - SS-SwinUnet
T2 - 2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024
AU - Ma, Bowei
AU - Qiu, Dehui
AU - Xiong, Ze
AU - Hu, Yulong
AU - Deng, Liguo
AU - Yuan, Huimei
AU - Wan, Xiaohua
AU - Zhang, Fa
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Precise segmentation of the pupil, iris, and sclera is critical for diagnosing and treating ocular diseases such as glaucoma, strabismus, and retinal disorders. However, the fine structural differences within the eye and the interference of complex backgrounds, especially with VR devices prone to reflections, tilts, distortions, and occlusions, present significant challenges. In this paper, we introduce SS-SwinUnet, a novel segmentation method that integrates Swin Transformer and knowledge distillation to achieve superior performance. Specifically, SS-SwinUnet balances feature transfer between the encoder and decoder, reducing redundancy and enhancing representation. Additionally, we incorporate a Boundary Difference over Union Loss to improve boundary segmentation accuracy. We also propose an eye modeling method that parameterizes segmentation results to optimize the semantic segmentation of ocular structures. We constructed the TongRenD dataset, comprising 400 VR-captured videos and 4,100 images, which, along with the TEyeD dataset, was used in our experiments. Results demonstrate that SS-SwinUnet significantly outperforms existing medical image segmentation methods across multiple datasets.
AB - Precise segmentation of the pupil, iris, and sclera is critical for diagnosing and treating ocular diseases such as glaucoma, strabismus, and retinal disorders. However, the fine structural differences within the eye and the interference of complex backgrounds, especially with VR devices prone to reflections, tilts, distortions, and occlusions, present significant challenges. In this paper, we introduce SS-SwinUnet, a novel segmentation method that integrates Swin Transformer and knowledge distillation to achieve superior performance. Specifically, SS-SwinUnet balances feature transfer between the encoder and decoder, reducing redundancy and enhancing representation. Additionally, we incorporate a Boundary Difference over Union Loss to improve boundary segmentation accuracy. We also propose an eye modeling method that parameterizes segmentation results to optimize the semantic segmentation of ocular structures. We constructed the TongRenD dataset, comprising 400 VR-captured videos and 4,100 images, which, along with the TEyeD dataset, was used in our experiments. Results demonstrate that SS-SwinUnet significantly outperforms existing medical image segmentation methods across multiple datasets.
KW - Eye modeling
KW - Feature extraction
KW - Knowledge distillation
KW - Ocular image segmentation
KW - Swin transformer
UR - http://www.scopus.com/inward/record.url?scp=85217280161&partnerID=8YFLogxK
U2 - 10.1109/BIBM62325.2024.10822559
DO - 10.1109/BIBM62325.2024.10822559
M3 - Conference contribution
AN - SCOPUS:85217280161
T3 - Proceedings - 2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024
SP - 2291
EP - 2296
BT - Proceedings - 2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024
A2 - Cannataro, Mario
A2 - Zheng, Huiru
A2 - Gao, Lin
A2 - Cheng, Jianlin
A2 - de Miranda, Joao Luis
A2 - Zumpano, Ester
A2 - Hu, Xiaohua
A2 - Cho, Young-Rae
A2 - Park, Taesung
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 3 December 2024 through 6 December 2024
ER -