Detecting glaucoma worsening using optical coherence tomography derived visual field estimates.

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Glaucoma progression is typically monitored using visual field (VF) tests.
• Optical coherence tomography (OCT) provides cross-sectional structural data.
• Estimating VF mean deviation (MD) from OCT aims to supplement VF testing.

Purpose of the Study:

• To develop and assess a machine learning (ML) model for converting OCT data to VF-MD estimates.
• To evaluate the ability of OCT-derived MD to detect longitudinal VF progression in glaucoma.

Main Methods:

• Trained an ML model on 70,575 paired OCT/VF datasets.
• Used a progression dataset of 4,044 eyes with ≥5 paired OCT/VFs.
• Calculated MD slopes using VF-MD and OCT-derived MD, comparing their ability to detect progression via Area Under the Curve (AUC).

Main Results:

• The ML model achieved a Mean Absolute Error (MAE) of 1.62 dB for OCT-MD estimates, outperforming previous models.
• AUC for MD slopes using partial or full OCT-MD substitution was significantly inferior to VF-MD alone.
• Supplementing VF-MD with OCT-MD did not improve AUC; an MAE ≤1.00 dB is needed for similar performance.

Conclusions:

• The developed ML model, while accurate, is currently insufficient for reliably detecting longitudinal VF progression in glaucoma.
• Future OCT-to-VF-MD models require improved prediction accuracy (MAE ≤1.00 dB) to become clinically useful for monitoring glaucoma worsening.