Glaucoma is the leading cause of irreversible blindness worldwide, affecting an estimated 80 million people. Unlike AMD and DR, glaucoma typically affects peripheral vision first — meaning patients may be completely unaware of the disease until significant optic nerve damage has occurred.

The hallmark of glaucoma is progressive excavation of the optic nerve head (ONH). The cup-to-disc ratio (CDR) quantifies this excavation: it expresses the diameter of the central cup (the pale, depressed zone of the disc) as a fraction of the total optic disc diameter. An elevated CDR suggests loss of rim tissue — the neuroretinal rim containing the retinal ganglion cell axons.

Why CDR is the primary screening metric

CDR measurement is standard practice in all ophthalmic examinations and is the most widely used single-variable glaucoma screening indicator. It is rapid to obtain from fundus photographs and requires no additional imaging equipment. However, CDR must always be interpreted in the context of disc size (large physiological discs can have high CDR without disease) and intraocular pressure measurements.

Limitations and clinical context

CDR alone cannot diagnose glaucoma. A suspicious CDR should prompt formal visual field testing (perimetry), optical coherence tomography (OCT) of the RNFL, and clinical measurement of intraocular pressure. The EyeMap CDR model is a triage tool to identify individuals warranting this workup, not a replacement for it.

CDR Interpretation Scale

0.30.50.60.8
0.3 – 0.5 Normal Majority of population. Routine recall unless other risk factors present.
0.5 – 0.6 Context-dependent Large but often physiological. Disc size critical. Recommend IOP measurement and repeat in 12 months.
> 0.6 Suspicious Possible early glaucomatous change. Refer for full glaucoma workup including perimetry, OCT, IOP.
⚠ Context-Sensitive Interpretation CDR interpretation requires knowledge of disc size. Macrodiscs (large physiological discs) may have a CDR of 0.7+ without any glaucomatous damage. The model output must be reviewed alongside clinical context by an optometrist or ophthalmologist.

Model architecture & validation

Input
Disc-centred fundus image
15°–45° field, optic disc centred
Output
Continuous CDR estimate
0.0–1.0 + segmentation overlay
Architecture
Segmentation + regression
U-Net disc/cup segmentation followed by ratio computation
Training data
REFUGE / DRISHTI
Benchmark glaucoma segmentation datasets
CDR agreement
±0.05 MAE
vs expert grader CDR measurement
Output overlay
Disc/cup map
Colour segmentation mask for clinician review