Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising

Xiangyang Yu; Yinan Wen; Haojin Li; Sanqian Li; Heng Li; Jiang Liu

doi:10.1007/978-981-92-3504-9_19

Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising

Xiangyang Yu
, Yinan Wen
, Haojin Li
, Sanqian Li^*
, Heng Li^*
, Jiang Liu^*

^*Corresponding author for this work

Southern University of Science and Technology
Jiangxi University of Finance and Economics
Shenzhen University of Advanced Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Intra-operative Optical Coherence Tomography (iOCT) is essential for ophthalmic surgery but is heavily degraded by speckle noise, while real-time acquisition prevents multi-frame averaging. Existing self-supervised Blind-Spot Networks (BSNs) suppress noise but often over-smooth direction-sensitive anatomical boundaries. We propose GE-BSN, a Geometric-Equivariant Blind-Spot Network for edge-sensitive iOCT denoising. GE-BSN embeds geometric equivariance into the asymmetric blind-spot mechanism through Geometric Prior-Guided Conv (GPG-Conv), enabling orientation-aware structural regularization. We further introduce Adaptive Anatomy-Preserving Fusion (AAP-Fusion) to balance strong background smoothing with high-fidelity boundary preservation. Experiments on real-world iOCT datasets show competitive denoising and improved structural fidelity for downstream retinal layer segmentation.

Original language	English
Title of host publication	Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings
Editors	De-Shuang Huang, Qinhu Zhang, Yijie Pan, Chuanlei Zhang, Wei Chen, Bo Li, Wenzheng Bao, Prashan Premaratne
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	230-240
Number of pages	11
ISBN (Print)	9789819235032
DOIs	https://doi.org/10.1007/978-981-92-3504-9_19
Publication status	Published - 2027
Externally published	Yes
Event	22nd International Conference on Intelligent Computing, ICIC 2026 - Toronto, Canada Duration: 22 Jul 2026 → 26 Jul 2026

Publication series

Name	Lecture Notes in Computer Science
Volume	16659 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	22nd International Conference on Intelligent Computing, ICIC 2026
Country/Territory	Canada
City	Toronto
Period	22/07/26 → 26/07/26

Keywords

Blind-Spot Network
Geometric Equivariance
OCT Denoising
Self-Supervised Learning

Access to Document

10.1007/978-981-92-3504-9_19

Cite this

Yu, X., Wen, Y., Li, H., Li, S., Li, H., & Liu, J. (2027). Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising. In D.-S. Huang, Q. Zhang, Y. Pan, C. Zhang, W. Chen, B. Li, W. Bao, & P. Premaratne (Eds.), Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings (pp. 230-240). (Lecture Notes in Computer Science; Vol. 16659 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-981-92-3504-9_19

Yu, Xiangyang ; Wen, Yinan ; Li, Haojin et al. / Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising. Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings. editor / De-Shuang Huang ; Qinhu Zhang ; Yijie Pan ; Chuanlei Zhang ; Wei Chen ; Bo Li ; Wenzheng Bao ; Prashan Premaratne. Springer Science and Business Media Deutschland GmbH, 2027. pp. 230-240 (Lecture Notes in Computer Science).

@inproceedings{c54fa14542074a53b527b1de01bc7cc6,

title = "Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising",

abstract = "Intra-operative Optical Coherence Tomography (iOCT) is essential for ophthalmic surgery but is heavily degraded by speckle noise, while real-time acquisition prevents multi-frame averaging. Existing self-supervised Blind-Spot Networks (BSNs) suppress noise but often over-smooth direction-sensitive anatomical boundaries. We propose GE-BSN, a Geometric-Equivariant Blind-Spot Network for edge-sensitive iOCT denoising. GE-BSN embeds geometric equivariance into the asymmetric blind-spot mechanism through Geometric Prior-Guided Conv (GPG-Conv), enabling orientation-aware structural regularization. We further introduce Adaptive Anatomy-Preserving Fusion (AAP-Fusion) to balance strong background smoothing with high-fidelity boundary preservation. Experiments on real-world iOCT datasets show competitive denoising and improved structural fidelity for downstream retinal layer segmentation.",

keywords = "Blind-Spot Network, Geometric Equivariance, OCT Denoising, Self-Supervised Learning",

author = "Xiangyang Yu and Yinan Wen and Haojin Li and Sanqian Li and Heng Li and Jiang Liu",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2027.; 22nd International Conference on Intelligent Computing, ICIC 2026 ; Conference date: 22-07-2026 Through 26-07-2026",

year = "2027",

doi = "10.1007/978-981-92-3504-9\_19",

language = "English",

isbn = "9789819235032",

series = "Lecture Notes in Computer Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "230--240",

editor = "De-Shuang Huang and Qinhu Zhang and Yijie Pan and Chuanlei Zhang and Wei Chen and Bo Li and Wenzheng Bao and Prashan Premaratne",

booktitle = "Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings",

address = "Germany",

}

Yu, X, Wen, Y, Li, H, Li, S, Li, H & Liu, J 2027, Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising. in D-S Huang, Q Zhang, Y Pan, C Zhang, W Chen, B Li, W Bao & P Premaratne (eds), Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings. Lecture Notes in Computer Science, vol. 16659 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 230-240, 22nd International Conference on Intelligent Computing, ICIC 2026, Toronto, Canada, 22/07/26. https://doi.org/10.1007/978-981-92-3504-9_19

Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising. / Yu, Xiangyang; Wen, Yinan; Li, Haojin et al.
Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings. ed. / De-Shuang Huang; Qinhu Zhang; Yijie Pan; Chuanlei Zhang; Wei Chen; Bo Li; Wenzheng Bao; Prashan Premaratne. Springer Science and Business Media Deutschland GmbH, 2027. p. 230-240 (Lecture Notes in Computer Science; Vol. 16659 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising

AU - Yu, Xiangyang

AU - Wen, Yinan

AU - Li, Haojin

AU - Li, Sanqian

AU - Li, Heng

AU - Liu, Jiang

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2027.

PY - 2027

Y1 - 2027

N2 - Intra-operative Optical Coherence Tomography (iOCT) is essential for ophthalmic surgery but is heavily degraded by speckle noise, while real-time acquisition prevents multi-frame averaging. Existing self-supervised Blind-Spot Networks (BSNs) suppress noise but often over-smooth direction-sensitive anatomical boundaries. We propose GE-BSN, a Geometric-Equivariant Blind-Spot Network for edge-sensitive iOCT denoising. GE-BSN embeds geometric equivariance into the asymmetric blind-spot mechanism through Geometric Prior-Guided Conv (GPG-Conv), enabling orientation-aware structural regularization. We further introduce Adaptive Anatomy-Preserving Fusion (AAP-Fusion) to balance strong background smoothing with high-fidelity boundary preservation. Experiments on real-world iOCT datasets show competitive denoising and improved structural fidelity for downstream retinal layer segmentation.

AB - Intra-operative Optical Coherence Tomography (iOCT) is essential for ophthalmic surgery but is heavily degraded by speckle noise, while real-time acquisition prevents multi-frame averaging. Existing self-supervised Blind-Spot Networks (BSNs) suppress noise but often over-smooth direction-sensitive anatomical boundaries. We propose GE-BSN, a Geometric-Equivariant Blind-Spot Network for edge-sensitive iOCT denoising. GE-BSN embeds geometric equivariance into the asymmetric blind-spot mechanism through Geometric Prior-Guided Conv (GPG-Conv), enabling orientation-aware structural regularization. We further introduce Adaptive Anatomy-Preserving Fusion (AAP-Fusion) to balance strong background smoothing with high-fidelity boundary preservation. Experiments on real-world iOCT datasets show competitive denoising and improved structural fidelity for downstream retinal layer segmentation.

KW - Blind-Spot Network

KW - Geometric Equivariance

KW - OCT Denoising

KW - Self-Supervised Learning

UR - https://www.scopus.com/pages/publications/105046169775

U2 - 10.1007/978-981-92-3504-9_19

DO - 10.1007/978-981-92-3504-9_19

M3 - Conference contribution

AN - SCOPUS:105046169775

SN - 9789819235032

T3 - Lecture Notes in Computer Science

SP - 230

EP - 240

BT - Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings

A2 - Huang, De-Shuang

A2 - Zhang, Qinhu

A2 - Pan, Yijie

A2 - Zhang, Chuanlei

A2 - Chen, Wei

A2 - Li, Bo

A2 - Bao, Wenzheng

A2 - Premaratne, Prashan

PB - Springer Science and Business Media Deutschland GmbH

T2 - 22nd International Conference on Intelligent Computing, ICIC 2026

Y2 - 22 July 2026 through 26 July 2026

ER -

Yu X, Wen Y, Li H, Li S, Li H, Liu J. Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising. In Huang DS, Zhang Q, Pan Y, Zhang C, Chen W, Li B, Bao W, Premaratne P, editors, Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings. Springer Science and Business Media Deutschland GmbH. 2027. p. 230-240. (Lecture Notes in Computer Science). doi: 10.1007/978-981-92-3504-9_19

Geometric-Equivariant Blind-Spot Network for Self-Supervised Intra-operative OCT Denoising

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this