A Comparative Study of Segmentation Techniques for Glaucoma Detection Using Fundus Images
VFAST Transactions on Software Engineering · 29.06.2026 · 14 (2) · 405-416
DOI: 10.21015/vtse.v14i2.2294 ↗Kurz gefasst
Fachartikel in der Zeitschrift VFAST Transactions on Software Engineering, online erschienen am 29.06.2026. Verfasst von Hashir Irfan und 2 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben.
Abstract (Originalsprache)
Glaucoma is a leading cause of irreversible blindness, and successful treatment largely depends on early and accurate diagnosis. A glaucoma classification approach using color fundus images is proposed in this study. Each image was preprocessed by resizing, applying Gaussian blurring to reduce noise, and enhancing contrast using CLAHE. Three segmentation methods were evaluated to isolate the optic disc: Otsu thresholding, elliptical boundary estimation with a fixed threshold of 185, and a U-Net–based deep learning model. While the U-Net provided the most precise optic disc segmentation, surprisingly, the Otsu-based segmentation yielded the best performance in downstream classification. This suggests that including regions surrounding the optic disc may provide additional discriminative information for distinguishing between glaucomatous and healthy eyes. Motivated by this observation, feature extraction was also performed directly on the full fundus images without restricting analysis to the segmented optic disc. This approach achieved an accuracy of 98.9%, demonstrating that peripheral regions around the optic disc contain meaningful information and should be considered in feature extraction strategies for glaucoma classification.
Zitieren
APA
Irfan, H., Suhail, Z., & Danish, H. M. (2026). A Comparative Study of Segmentation Techniques for Glaucoma Detection Using Fundus Images. VFAST Transactions on Software Engineering, 14(2), 405-416. https://doi.org/10.21015/vtse.v14i2.2294
Vancouver
Irfan H, Suhail Z, Danish HM. A Comparative Study of Segmentation Techniques for Glaucoma Detection Using Fundus Images. VFAST Transactions on Software Engineering. 2026;14(2):405-416. doi:10.21015/vtse.v14i2.2294
BibTeX
@article{Irfan2026,
author = {Irfan, Hashir and Suhail, Zobia and Danish, Hafiz Muhammad},
title = {A Comparative Study of Segmentation Techniques for Glaucoma Detection Using Fundus Images},
journal = {VFAST Transactions on Software Engineering},
year = {2026},
volume = {14},
number = {2},
pages = {405--416},
doi = {10.21015/vtse.v14i2.2294}
}
EUVIMED