Concepts

The science behind GlaucomAI

Every idea the pipeline relies on, from eye pressure to AI metrics. Each concept has its own page with a diagram, and most diagrams respond to the controls under them. Shorter terms are in the glossary, and every page lists the sources it cites.

What glaucoma is

An optic nerve disease: ganglion-cell axons are lost, the disc cups and blind spots appear.

Eye pressure (IOP)

How aqueous humour sets eye pressure, and why 21 mmHg is a statistical line, not a safe limit.

Normal-tension glaucoma
interactive

Glaucoma at normal pressure: common, and a hint that blood supply matters.

Optic disc and CDR
interactive

Rim, cup and the vertical cup-to-disc ratio, with the ISNT rule.

Photo to CDR with AI

How GlaucomAI's two U-Nets find the disc and outline disc and cup.

OCT imaging
interactive

Cross-sections of the retina: cup depth and the nerve fibre layer.

Perfusion pressure
interactive

Blood pressure minus eye pressure: the push that feeds the optic nerve.

Autoregulation
interactive

How vessels keep blood flow steady, and what happens when they can't.

Flicker test
interactive

Flickering light makes retinal vessels widen; in glaucoma they widen less.

Visual fields, progression

Arc-shaped field defects, mean deviation and the rate of loss in dB per year.

Screening, ROC, AUC
interactive

Risk scores, thresholds, sensitivity, specificity and the area under the ROC curve.

Training vs validation
interactive

External validation, and why both eyes of a patient stay on one side of a split.

Vascular phenotypes
interactive

GlaucomAI's labels: pressure-dominant, vascular-risk, mixed, and flicker-confirmed.

Glossary
interactive

40 shorter terms, from OD/OS to pseudonymisation, searchable by category.

References

Every source cited on the concept pages that are switched on, with links.