<![CDATA[The Cup-to-Disc Ratio (CDR) is widely used as a diagnostic indicator for glaucoma, although variations and irregularities can influence its accuracy in the Optic Disc (OD) and Optic Cup (OC). To overcome this challenge, automated image segmentation is used. However, image segmentation is challenged by image blurriness, noise, and uneven illumination, which can affect segmentation quality and increase the risk of misdiagnosis. To address these challenges, this study applies a combined Flip-Gamma Augmentation and SegFormer approach for OD and OC segmentation. Flip-Gamma augmentation increases image diversity and improves image quality by adjusting brightness and contrast. Meanwhile, the SegFormer uses a Transformer-based backbone and efficiently extracts multi-scale features to enhance segmentation performance. Experimental results on the Drishti-GS dataset show that applying Flip-Gamma (δ = 0.8, 0.9, 1.1, 1.2) is associated with improved segmentation performance across all classes, with sensitivity (90-99%), DSC (90-99%), IoU (82-99%), and ROC (94-99%), indicating consistent segmentation of OD, OC and background regions. Furthermore, a one-sided Mann-Whitney U test indicates differences in performance compared to other augmentation methods. These findings suggest that the proposed augmentation strategy is beneficial for segmentation on the Drishti-GS dataset. However, further validation on larger and more diverse datasets is required to assess generalizability.]]>
Copyrights © 2026
