AZmed publishes five years of X-ray AI results from CHU de Reims

Most imaging AI numbers come from the company selling the software. This set does not. AZmed points to a hospital study run without it, which is the kind of evidence a radiology buyer can actually lean on.

By Atlas ... AI Editorial Collaborator

What AZmed Reported About Five Years at CHU de Reims

AZmed said on 7 October 2026 that the radiology department at CHU de Reims has used AZtrauma for fracture detection and AZchest for chest X-rays since 2021. The company cites a study in Pediatric Radiology covering 366 children under 18, which it says was run independently of AZmed and compared several fracture detection algorithms, reporting 93.6 percent specificity for Rayvolve and, for less experienced residents, sensitivity rising from 87.0 to 91.3 percent and specificity from 88.2 to 93.2 percent with AI help. AZmed separately cites a Radiography evaluation of 195,706 musculoskeletal X-rays, reporting an area under the curve of 0.983 for AZtrauma and 0.978 for AZchest.

Why the AZmed Reims Results Matter to Radiology Buyers

The part worth reading twice is who ran the study. A hospital evaluation published in a journal and conducted without the vendor is stronger evidence than a company data sheet, and the Reims numbers show the biggest gain going to the least experienced readers. That is a useful way to think about where this kind of tool pays for itself. It is still one hospital and 366 children, so it describes that department rather than a promise for a busy adult trauma service.

Where the AZmed Update Comes From

AZmed is the original record behind this update. It tells us what the company published. This brief adds the market context and the method we would use to test the development against other evidence.

AZmed original source.

How We Would Research the AZmed Reims Results

The source gives us the starting point. This is how we would build the next layer of research around it.

We would start with the Pediatric Radiology paper itself and check that the independence claim holds, including funding and author affiliations.
Then we would look at how the gain was split between experienced and less experienced readers, because that decides whether the tool is a safety net or a second opinion.
We would ask AZmed which software version was running during the five years at Reims, since an algorithm that has been recleared several times is not the one the study measured.