JMIR Publications

ABSTRACT

Background:

Optic disc tilt is a morphological change in myopic eyes that complicates clinical interpretation and artificial intelligence (AI)-based analysis of fundus images. Accurate detection of optic disc tilt is necessary to avoid misinterpretation of disc morphology and enhance diagnostic reliability across different disease types.

Objective:

This study developed and externally validated an end-to-end AI-based pipeline for optic disc segmentation and quantitative tilt classification in color fundus photographs (CFPs), offering an objective alternative to manual segmentation and subjective clinical assessments.

Methods:

We trained a nnU-Net-based optic disc segmentation model on the Standardized Multi-Channel Dataset for Glaucoma (SMDG; 3,103 CFPs) and externally validated it on the Samsung Medical Center dataset (2,448 CFPs; 1,958 patients). Tilt was classified using the ratio of the long-distance diameter to short-distance diameter, with a ratio ≥ 1.3 indicating tilt. Segmentation performance was evaluated using the dice similarity coefficient (DSC), intersection over union (IoU), and pixel accuracy on the SMDG and the clinical acceptance rate via expert review.

Results:

Using the SMDG, nnU-Net achieved outstanding performance (mean ± SD: DSC, 0.961 ± 0.055; IoU, 0.927 ± 0.057) across eight datasets. With the SMC dataset, expert review showed a mean clinical acceptance rate of 98.61% across disease types, ranging from 86.40% (edema) to 99.59% (pallor). Tilt was detected in 7.5% (186/2,448) of images, with rates of 9.7% (normal), 3.9% (glaucoma), 7.8% (pallor), and 14.2% (edema). Segmentation errors occurred in 1.4% (34/2,448) of cases, mainly due to edema-related swelling, peripapillary atrophy, and vessel confusion.

Conclusions:

Our pipeline provides objective and reproducible detection of optic disc tilt on CFPs, with strong generalization to clinical images. Replacing manual segmentation and subjective assessments, the pipeline supports tilt-aware AI diagnostics and scalable screening for myopia-related conditions, with future refinements needed for edema-related challenges.