AZmed wins 6th FDA clearance, expanding AZchest on chest X-rays

Paris medtech AZmed said the FDA cleared AZchest to detect and localize pneumothorax, pleural effusion, consolidation, and increased cardiothoracic ratio on frontal chest radiographs — its sixth U.S. clearance. With earlier lung-nodule detection, AZchest now covers five chest findings on one image; the broader Rayvolve suite reaches eight cleared findings across musculoskeletal and chest X-rays.

A same-day company primary plus an FDA 510(k) record expands a chest X-ray AI helper to findings that change urgent care — collapsed lung, fluid, consolidation, and heart-size ratio — with multi-reader evidence the company says improves accuracy and speed.

AZmed, a Paris medical-technology company, announced its sixth U.S. Food and Drug Administration clearance for AZchest, AI software that looks for chest problems on frontal chest X-rays. That 15 Sep 2026 company news post is the filing event. These are company claims.

Newly cleared findings, as the company lists them: pneumothorax, pleural effusion, consolidation, and increased cardiothoracic ratio. The software is cleared to detect and localize those findings — find them and mark where they are. In plain English: pneumothorax is an air leak that can collapse a lung; pleural effusion is fluid around the lung; consolidation is lung tissue filled in, often an infection or pneumonia pattern; cardiothoracic ratio is how wide the heart looks compared with the chest. This is assist software, not a diagnosis on its own.

Together with lung-nodule detection, which the company says was cleared in March 2025, AZchest now covers five findings on a single image. The broader Rayvolve suite, the company says, now reaches eight FDA-cleared findings across musculoskeletal X-rays — bones, joints, and muscles — and chest X-rays. File five-on-one-image and eight-across-MSK-and-chest as AZmed’s.

The FDA 510(k) database lists Device Name “Rayvolve AZchest,” 510(k) number K261378, applicant AZmed of Paris, decision date 28 Aug 2026, decision Substantially Equivalent (SESE). A 510(k) is the common FDA path that clears a device as substantially equivalent to something already on the market. The company’s public announce of this expansion is 15 Sep 2026 — that announce date is the news wire; 28 Aug 2026 is the FDA decision on the record.

Standalone study, company-attributed: 8,453 per-finding evaluations; area under the curve, or AUC, of 0.94 to 0.98 for all four findings. AUC is a detection-accuracy score from 0 to 1; closer to 1.0 is better. Per-image sensitivity 90.8% to 97.0%; specificity 87.9% to 92.1%. Sensitivity is how often the tool catches a finding that is there; specificity is how often it stays quiet when the finding is not. The company post does not name a journal for this standalone run.

Multi-reader, multi-case study, still the company’s: 1,489 per-finding evaluations; 20 readers, including 10 thoracic radiologists — chest-imaging specialists. Readers first worked without the tool, then with AZchest after a one-month washout. Reader AUC rose from 0.88–0.89 unaided to 0.91–0.94 aided; the company says sensitivity and specificity also improved. Thoracic radiologists were up to 15.9% faster; readers from other specialties up to 31.0% faster.

SimonMed, a U.S. Outpatient imaging provider, provided the clinical data behind this clearance, the company said. The relationship dates to 2023. The company also says SimonMed data supported an earlier pediatric-fracture clearance.

Julien Vidal, CEO of AZmed, quoted on the company post: “Radiologists should not need a different AI tool for every finding on the same image.” And: “We are building one suite that covers the X-ray, and every step of it carries strong clinical evidence.” They are not an independent product review, and they are not a claim the tool replaces radiologists.

Company about box: founded 2018; European MedTech; Rayvolve tools described as FDA-cleared and CE-marked; used by more than 2,500 healthcare facilities worldwide.

What this is not: AZchest is Class II computer-aided detection / assist software — a helper for clinicians reviewing chest X-rays, not autonomous diagnosis, not a drug approval, and not a claim it replaces radiologists or cuts deaths. Extra 510(k) numbers beyond K261378 are not invented here.