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.
An optic nerve disease: ganglion-cell axons are lost, the disc cups and blind spots appear.
How aqueous humour sets eye pressure, and why 21 mmHg is a statistical line, not a safe limit.
Glaucoma at normal pressure: common, and a hint that blood supply matters.
Rim, cup and the vertical cup-to-disc ratio, with the ISNT rule.
How GlaucomAI's two U-Nets find the disc and outline disc and cup.
Cross-sections of the retina: cup depth and the nerve fibre layer.
Blood pressure minus eye pressure: the push that feeds the optic nerve.
How vessels keep blood flow steady, and what happens when they can't.
Flickering light makes retinal vessels widen; in glaucoma they widen less.
Arc-shaped field defects, mean deviation and the rate of loss in dB per year.
Risk scores, thresholds, sensitivity, specificity and the area under the ROC curve.
External validation, and why both eyes of a patient stay on one side of a split.
GlaucomAI's labels: pressure-dominant, vascular-risk, mixed, and flicker-confirmed.
40 shorter terms, from OD/OS to pseudonymisation, searchable by category.
Every source cited on the concept pages that are switched on, with links.