SS-SwinUnet: A Distillation Method of Swin Transformer for Superior Ocular Image Segmentation

Bowei Ma, Dehui Qiu*, Ze Xiong, Yulong Hu, Liguo Deng, Huimei Yuan, Xiaohua Wan*, Fa Zhang*

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

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.

Original languageEnglish
Title of host publicationProceedings - 2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024
EditorsMario Cannataro, Huiru Zheng, Lin Gao, Jianlin Cheng, Joao Luis de Miranda, Ester Zumpano, Xiaohua Hu, Young-Rae Cho, Taesung Park
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages2291-2296
Number of pages6
ISBN (Electronic)9798350386226
DOIs
Publication statusPublished - 2024
Event2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024 - Lisbon, Portugal
Duration: 3 Dec 20246 Dec 2024

Publication series

NameProceedings - 2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

Conference

Conference2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024
Country/TerritoryPortugal
CityLisbon
Period3/12/246/12/24

Keywords

  • Eye modeling
  • Feature extraction
  • Knowledge distillation
  • Ocular image segmentation
  • Swin transformer

Fingerprint

Dive into the research topics of 'SS-SwinUnet: A Distillation Method of Swin Transformer for Superior Ocular Image Segmentation'. Together they form a unique fingerprint.

Cite this