GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography

Abstract Dust and sensor noise often create artificial spots in fundus photography, and clinicians may occasionally misinterpret them as pathological signs such as microaneurysms. Reliable computer-aided diagnosis depends on accurately identifying and segmenting such artifacts. However, producing pi...

Full description

Saved in:
Bibliographic Details
Main Authors: Jaehan Joo, Hunyoul Lee, Suk Chan Kim
Format: Article
Language:English
Published: Nature Portfolio 2025-07-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-07077-4
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849763798420815872
author Jaehan Joo
Hunyoul Lee
Suk Chan Kim
author_facet Jaehan Joo
Hunyoul Lee
Suk Chan Kim
author_sort Jaehan Joo
collection DOAJ
description Abstract Dust and sensor noise often create artificial spots in fundus photography, and clinicians may occasionally misinterpret them as pathological signs such as microaneurysms. Reliable computer-aided diagnosis depends on accurately identifying and segmenting such artifacts. However, producing pixel-level annotations for these subtle structures remains labor-intensive and challenging to scale. We propose GMS-JIGNet, a self-supervised segmentation framework based on guided multi-scale jigsaw puzzles and contrastive learning, to address this issue. The method learns spatially-aware representations from unlabeled data by solving jigsaw puzzles across multiple resolutions while selectively injecting positional hints for uninformative regions. The downstream segmentation model receives these representations and uses the ViT encoders from the pretext task as fixed feature extractors and a lightweight FPN decoder. Experimental results on a large-scale fundus dataset show that our proposed model achieves state-of-the-art performance across various metrics, including IoU, DICE, and SSIM, even when trained with only a few labeled images. Moreover, we conducted ablation studies to assess how well our architecture performs under different training hyperparameter setups. The results support the effectiveness of guided self-supervised learning in medical image segmentation and suggest its strong potential for clinical use, especially in settings with limited labeled data.
format Article
id doaj-art-39178e7a615e4b6e96e2f5e96cbfb0d1
institution DOAJ
issn 2045-2322
language English
publishDate 2025-07-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-39178e7a615e4b6e96e2f5e96cbfb0d12025-08-20T03:05:18ZengNature PortfolioScientific Reports2045-23222025-07-0115111710.1038/s41598-025-07077-4GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photographyJaehan Joo0Hunyoul Lee1Suk Chan Kim2Electric and Electronics Engineering, Pusan National UniversityElectric and Electronics Engineering, Pusan National UniversityElectric and Electronics Engineering, Pusan National UniversityAbstract Dust and sensor noise often create artificial spots in fundus photography, and clinicians may occasionally misinterpret them as pathological signs such as microaneurysms. Reliable computer-aided diagnosis depends on accurately identifying and segmenting such artifacts. However, producing pixel-level annotations for these subtle structures remains labor-intensive and challenging to scale. We propose GMS-JIGNet, a self-supervised segmentation framework based on guided multi-scale jigsaw puzzles and contrastive learning, to address this issue. The method learns spatially-aware representations from unlabeled data by solving jigsaw puzzles across multiple resolutions while selectively injecting positional hints for uninformative regions. The downstream segmentation model receives these representations and uses the ViT encoders from the pretext task as fixed feature extractors and a lightweight FPN decoder. Experimental results on a large-scale fundus dataset show that our proposed model achieves state-of-the-art performance across various metrics, including IoU, DICE, and SSIM, even when trained with only a few labeled images. Moreover, we conducted ablation studies to assess how well our architecture performs under different training hyperparameter setups. The results support the effectiveness of guided self-supervised learning in medical image segmentation and suggest its strong potential for clinical use, especially in settings with limited labeled data.https://doi.org/10.1038/s41598-025-07077-4Self-supervised learningFundus photographyJigsaw puzzleArtificial spotSegmentation
spellingShingle Jaehan Joo
Hunyoul Lee
Suk Chan Kim
GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
Scientific Reports
Self-supervised learning
Fundus photography
Jigsaw puzzle
Artificial spot
Segmentation
title GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
title_full GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
title_fullStr GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
title_full_unstemmed GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
title_short GMS-JIGNet: guided multi-scale jigsaw puzzles for self-supervised artificial spot segmentation in fundus photography
title_sort gms jignet guided multi scale jigsaw puzzles for self supervised artificial spot segmentation in fundus photography
topic Self-supervised learning
Fundus photography
Jigsaw puzzle
Artificial spot
Segmentation
url https://doi.org/10.1038/s41598-025-07077-4
work_keys_str_mv AT jaehanjoo gmsjignetguidedmultiscalejigsawpuzzlesforselfsupervisedartificialspotsegmentationinfundusphotography
AT hunyoullee gmsjignetguidedmultiscalejigsawpuzzlesforselfsupervisedartificialspotsegmentationinfundusphotography
AT sukchankim gmsjignetguidedmultiscalejigsawpuzzlesforselfsupervisedartificialspotsegmentationinfundusphotography