CorneaDx brings image-quality scoring, highlighted corneal findings, and structured decision support directly onto the live slit-lamp feed — so the ophthalmologist stays in command while every capture becomes consistent, explainable, and clinically useful.
Today's anterior-segment workflow depends on a single clinician's eyes, attention, and recall — and the most useful information is rarely captured in a structured, reviewable form.
Subtle corneal signs — early infiltrates, faint scars, epithelial changes — are easy to overlook, especially under time pressure or in a busy OPD.
Hand-written notes vary between doctors, between clinics, and between visits. Patterns across patients are hard to surface for review or audit.
Trainees rarely see consistent annotated captures of the same case over time. There is no scalable way to build a teaching corpus from real cases.
Referrals and second opinions depend on photos taken in passing — and rarely with the quality, angles, or annotation needed for confident review.
CorneaDx runs as a clinician-facing overlay on the slit-lamp live feed. It does the quality control and the bookkeeping so the ophthalmologist can keep their eyes on the patient.
The interface shows three things in one view: the live slit-lamp image, the AI's findings pinned to the same image, and a structured diagnosis card the doctor can edit, accept, or reject before sign-off.
Every capability listed below is built around one principle: the clinician stays in command, and the AI does the parts that computers do better — consistently, at speed, and without forgetting.
Connects to the slit-lamp live feed over a standard capture path. No new hardware required for most existing slit lamps — capture happens at clinical resolution, not downscaled.
Every frame is scored on focus, illumination, centration, lid遮挡, and media clarity. A capture is only marked "ready for review" when it crosses the clinical-quality threshold.
Findings are pinned to coordinates with class, confidence, and click-through evidence. The doctor always sees why the model said what it said — and can override it.
Output is not a paragraph — it is fields: location, depth, laterality, severity, suspected diagnosis, advice, follow-up. Each field is editable, and rejections feed the training set.
Drops into a single-room clinic or a multi-doctor hospital corridor. Case records export as a referral-ready PDF with image, findings, and the doctor's signed diagnosis.
Runs on-device first; only review-quality captures leave the room, and only with explicit consent. Built to align with India's DPDP Act for healthcare data.
CorneaDx turns the slit-lamp encounter into a structured, reviewable, and teachable artefact — without slowing the clinic down or taking the clinician out of the decision.
Fewer missed findings. Consistent image quality on every capture. Less time writing notes, more time on the patient. A referral packet that builds itself.
Standardised records across doctors. Faster audits. A growing, in-house corpus of clean, annotated cases — without a separate annotation project.
A teaching library built from real cases the department has already seen — with the senior's edits preserved as the ground truth.
A clearer explanation of what the doctor saw, what it might mean, and what happens next — because the doctor can show, not just tell.
If you run a single-room clinic, a multi-doctor hospital, or an academic department — we'd like to show you what the module looks like on your own slit lamp, with your own cases.