Association between visual field damage and corneal structural parameters.

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Predicting glaucoma progression and visual field loss is crucial for timely intervention.
• While central corneal thickness (CCT) and corneal hysteresis are known predictors, other corneal parameters' roles are less understood.

Purpose of the Study:

• To investigate the association between corneal shape, elevation, and thickness parameters and visual field damage in glaucoma patients.
• To evaluate the efficacy of machine learning models in predicting visual field abnormalities and severity based on corneal structural data.

Main Methods:

• Utilized anterior segment optical coherence tomography (OCT) for corneal topography, pachymetry, and elevation imaging.
• Collected visual field data using standard automated perimetry (24-2 Swedish Interactive Threshold Algorithm).
• Employed ensemble machine learning (bagged trees) and tree-based classifiers, with tenfold cross-validation and AUC for evaluation.

Main Results:

• Machine learning models accurately detected visual field abnormality from corneal parameters with an AUC of 0.83.
• A tree-based classifier identified four visual field severity levels using corneal parameters, achieving an AUC of 0.74.
• Average mean deviation was -8.0 dB and average CCT was 513.1 µm.

Conclusions:

• Corneal shape and elevation parameters, in addition to CCT, show potential as predictors of glaucoma-induced visual functional loss.
• Machine learning effectively leverages corneal structural data to identify and classify visual field damage.
• These findings may enhance glaucoma risk assessment and monitoring strategies.