A Structure-Function Dataset With Known Rates of Glaucomatous Progression.

Area of Science:

• Ophthalmology
• Medical Imaging
• Data Science

Background:

• Glaucoma is a leading cause of irreversible blindness.
• Accurate assessment of glaucoma progression is crucial for timely intervention.
• Existing datasets may lack ground-truth progression rates for method validation.

Purpose of the Study:

• To introduce the Lions Eye Institute Structure-Function Dataset (LEI-SFD).
• To provide an open, synthetic dataset of structural and functional eye measurements.
• To enable the assessment of glaucoma progression detection methods using known rates.

Main Methods:

• The LEI-SFD comprises 10 visits over 5 years of synthetic static automated perimetry (SAP) and circumpapillary retinal nerve fiber layer (cpRNFL) thickness data.
• Data is seeded with curated clinical measurements from the Lions Eye Institute, Perth, Australia.
• Measurement noise is incorporated based on existing literature to simulate real-world variability.

Main Results:

• The dataset includes 162 eyes with a mean baseline deviation of -2.6 dB and mean cpRNFL thickness of 77.7 microns.
• On average, 20.8 SAP locations per eye show progression at a mean rate of -0.6 dB/year.
• Permutation of Pointwise Linear Regression (PoPLR) on the synthetic data yields classification results comparable to those from clinical data.

Conclusions:

• The LEI-SFD offers longitudinal, linked structural and functional data with ground-truth progression rates.
• Synthetic generation based on real clinical data preserves structure-function relationships.
• This benchmark dataset facilitates the evaluation of glaucoma progression assessment tools.