CLINICAL EVIDENCE & VALIDATION
A decision support tool is only worth the evidence behind it. This page collects, in one place, where RLES AI rules come from, who checked them, what has been measured so far in the field, and — just as importantly — what has not yet been validated.
THE EVIDENCE BASE IN NUMBERS
Every clinical page on this site carries its own reference list with full citations, DOIs and PubMed identifiers. Across the platform:
Where an exact identifier could not be verified, the page links to a literature search rather than inventing a citation. This is a stated policy, not an accident — see the Editorial Policy.
HOW A RULE ENTERS THE PLATFORM
No threshold reaches a surgeon's report without passing through the same sequence. The point of writing it down is that it can be audited.
WHAT THE FIELD DATA SHOWS SO FAR
RLES AI has been in clinical testing since January 2026. The figures on the home page are read directly from the running service and are updated automatically — surgeon count, cases analysed, and the proportion of reports raising a yellow or a red flag. They are counts of system behaviour, and that is all they claim to be.
A flag rate is not an outcome measure. It records how often the safety engine raised a concern — not how often that concern was correct, and not how often the surgeon agreed. RLES AI does not yet publish sensitivity, specificity or concordance figures, because the prospective data needed to calculate them honestly does not exist yet. When it does, it will be published here with its methodology.
WHO REVIEWED WHAT
Clinical review is attributed, not implied. Each clinical page names the surgeons who reviewed it in the “Medically reviewed by” line beneath the headline, and the same reviewers appear in the page's structured data so that the human-readable and machine-readable claims match. The full board, with institutions, specialties and links to their publications, is on the About page.
DEVICE DOCUMENTATION
Laser and microkeratome parameters are treated as manufacturer facts, not clinical opinion. Published nomograms and instructions for use are loaded per platform; a device or a mode that is not in the catalogue is simply not offered. The supported platforms and the reasoning behind this are set out on the Supported Devices page.
PLATFORM VERSION HISTORY
RLES AI has been in development since February 2025; the first working version was released four months later. Versions are listed newest first, and each entry names what the release added, not how it performed.
CURRENT LIMITATIONS
- RLES AI is in clinical testing. It is not a certified medical device and carries no regulatory clearance for autonomous clinical use
- No prospective validation study has been completed. Agreement with expert review has not been quantified
- Field figures come from a pilot population of collaborating surgeons and are not a random or representative sample
- The platform does not calculate IOL power, does not recommend a lens brand or model, and does not overrule a surgeon
- Modules that depend on image analysis are disabled in the US Edition pending regulatory submission
- Every report is a decision support document. The clinical decision, and the responsibility for it, remain with the treating surgeon
WHAT COMES NEXT
The validation work now in preparation is a retrospective review of flagged versus unflagged cases against surgeon judgement, and a prospective comparison on ectasia screening. Results — including negative results — will be published on this page with their methods, sample sizes and dates, whether or not they favour the platform.
RLES AI runs as a mobile interface on your phone — nothing to install from a store, reports in minutes. The final clinical decision always rests with the surgeon.
Request Information