Press & partners

The facts, stated plainly.

Everything a journalist, funder, researcher or partner needs — including what GeoRadX cannot yet do.

Fact sheet

At a glance

What it is
A diagnostic-support tool that reads chest X-rays for signs of disease and flags regions for a clinician's review.
What it detects
Pleural effusion, consolidation, cardiomegaly. Three pathologies — stated openly, not more.
Regulatory status
PENDING  Awaiting NAFDAC clearance as a diagnostic-support medical device. Not cleared for clinical use. Expected December 2026.
Clinical deployment
NONE YET  It has never been used on a real patient. There are no field results, because there has been no field.
The architecture
DRF-MWU — a recursive field engine that resolves an image by converging on its spatial structure rather than classifying it against labels. Internals are proprietary and patent-pending.
Model size
Thousands of parameters, not billions. Small enough to run offline on an ordinary clinic computer — no GPU, no cloud, no internet.
Patient data
Images are processed on-device. Patient-identifying data never leaves the clinic. Contributed images are de-identified on the device before transmission.
How it learns
Exclusively from licensed clinicians who review anonymised images through a verified reviewer network. Not scraped from the internet.
Business model
A single meter: clinics pay per reading. No subscriptions, no per-seat fees, no paid modules. Small and rural clinics use it free, at the same accuracy.
Built by
OmegaRS — a deep-tech company specialising in edge deployment and lean AI architectures. The underlying theory originated at the Institute for Meta-Sciences, a non-profit research organisation.
Team
SOLO  Architecture, engine, product and company built by one researcher, working from Nigeria.
First market
Nigeria and West Africa, where the radiologist shortage is most acute — followed by other under-served regions.
For accuracy in reporting

What may — and may not — be said about GeoRadX

Medical AI is routinely over-reported. If you are writing about us, these are the boundaries we hold ourselves to, and we ask that coverage hold them too.

Accurate to say

  • "A diagnostic-support tool that flags possible findings for a clinician."
  • "Runs offline on ordinary clinic hardware, without internet or a GPU."
  • "Built on a novel geometric architecture rather than a large label-trained model."
  • "Pending regulatory approval; not yet used in clinical practice."
  • "Trained on labels from licensed clinicians."

Not accurate to say

  • "AI that diagnoses disease." It does not diagnose.
  • "Replaces radiologists." It is legally forbidden from operating alone.
  • "Approved" or "in use in clinics." Neither is true yet.
  • Any accuracy figure not published by us. We have not released one.
  • "Reads any chest condition." It detects three, and says so.
Get in touch

Who you are

Press

Interviews, background, and technical detail. We will tell you what the system cannot do as readily as what it can.

press@georadx.com

Funders & investors

Grant funding sustains the reviewer network and the free tier. Investment supports the engine and edge platform.

invest@georadx.com

Research & clinical partners

Hospitals, radiology departments and researchers interested in validation studies or collaboration.

research@georadx.com
Media kit

Assets

Logo & wordmark
SVG and PNG, light and dark
Download →
Product screenshots
The reading, the breakdown, and follow-up progression
Download →
One-page fact sheet
PDF · the table above, printable
Download →

Ask us the hard questions.

We would rather answer a sceptical question well than a friendly one easily. If something here does not add up, tell us.