Automated quantification of central corneal epithelial thickness for the diagnosis of limbal stem cell deficiency.
Publication Year:
2026
PubMed ID:
42832607
Funding Grants:
Public Summary:
Limbal stem cell deficiency (LSCD) is a serious eye condition that damages the cornea, the clear outer surface of the eye, and can lead to vision loss. Measuring the thickness of the cornea’s outermost layer is important for diagnosing LSCD and assessing disease severity. However, current automated tools often produce inaccurate measurements, particularly in patients with advanced disease, requiring time-consuming manual evaluation by specialists.
This study developed AI-CE, an artificial intelligence (AI)-based tool that automatically measures corneal epithelial thickness using noninvasive eye scans. The model was evaluated using images from 134 eyes representing healthy individuals and patients with different stages of LSCD. AI-CE demonstrated significantly greater accuracy than existing commercial software, reducing measurement errors by more than 50%. These findings were confirmed in an independent dataset. AI-CE also successfully distinguished between different stages of LSCD, with results comparable to expert manual measurements.
By providing faster, more accurate, and consistent measurements, AI-CE has the potential to improve LSCD diagnosis, disease monitoring, and treatment evaluation. This technology could reduce the burden on clinicians and support the development and assessment of new therapies aimed at restoring vision in patients with LSCD.
Scientific Abstract:
Limbal Stem Cell Deficiency (LSCD) is a corneal disorder characterized by conjunctivalization of the cornea. Central corneal epithelial thickness, measured using Anterior Segment Optical Coherence Tomography (AS-OCT), is a biomarker for LSCD. Manual measurement of epithelial thickness remains the gold standard for diagnosis because existing automated commercial software is unreliable in severe cases. Standard machine learning approaches have limited applicability for automating this measurement due to the difficulty of distinguishing corneal epithelium from scar tissue, and imprecise clinical labels. The goal of this study is to develop a deep learning pipeline to automate measurement of central epithelial thickness in AS-OCT images of eyes with LSCD. To accomplish this goal, we developed AI-CE, a model based on Multi-scale U-Net (MSU-Net) trained to accurately delineate the epithelial layer. The primary dataset comprised of 134 patient eyes, including 47 healthy normal eyes (35.0%), 34 with Stage 1 (25.4%), 34 with Stage 2 (25.4%), and 19 with Stage 3 (14.2%) LSCD. Performance was evaluated by calculating the average mean absolute error between model segmentations and manually measured clinical assessments. AI-CE model performance was compared against a commercially available segmentation software to demonstrate its improvement over the current state-of-the-art. The proposed model achieved a mean absolute error of 5.10 mum, significantly improving (p < 0.01) over clinical software (12.20 mum). This result was verified in a secondary retrospective dataset which similarly showed significant improvement over commercial software. We also demonstrate that clinical epithelial thickness values provided by AI-CE can distinguish between LSCD stages similarly to manual evaluations.