qXR

Performance Metrics

Clinical Evidence

qXR's evidence base is notable for two things: it is large, and much of it is independent of the manufacturer. The foundational diagnostic-accuracy study (Nash, Kik, Qure-independent investigators; Scientific Reports 2020;10:210) was a retrospective case-control evaluation at a tertiary Indian hospital using microbiologically-confirmed pulmonary tuberculosis as the reference standard. Across 317 cases and 612 controls, qXR achieved an AUC of 0.81 (95% CI 0.78-0.84) and, at the operating point maximising both, 71% sensitivity and 80% specificity — outperforming radiologists on sensitivity at matched specificity. Independent blinded head-to-head work followed: an evaluation of 12 commercial AI solutions for tuberculosis (Scientific Reports 2021) placed qXR in the top tier, and a South African tuberculosis-prevalence survey published in Lancet Digital Health 2024 externally validated 12 CAD products against microbiological evidence. Two randomised controlled trials anchor the higher-quality end of the evidence. A multicentre prospective RCT in Korean respiratory outpatient clinics (Lee HW et al., Annals of the American Thoracic Society 2023;20:660-667; 162 intervention vs 161 control patients) found qXR assistance significantly raised the area under the curve for non-radiologist physicians detecting lung lesions — though it did not change their clinical decisions. The larger and more recent LungIMPACT trial (Nature Medicine 2026), randomised across five NHS trusts and 93,326 chest X-rays, tested immediate AI prioritisation of primary-care chest X-rays and found no significant reduction in time to CT (median 53 days in both arms) or to lung-cancer diagnosis (44 vs 46 days). The authors concluded that prioritisation alone, without pathway redesign, is unlikely to accelerate the diagnostic pathway — an important, candid negative result for one of the product's headline use cases. Prospective and real-world cohorts round out the picture. A prospective study in Lima, Peru (PLOS Global Public Health 2024; 1,006 patients) reported very high tuberculosis-triage sensitivity (0.91, 95% CI 0.81-0.97 against culture) but low specificity (0.32) — the classic high-sensitivity-low-specificity triage trade-off. A prospective multicentre quality-improvement study across 35 Indian centres (Diagnostics 2022; 65,604 radiographs) and a post-deployment evaluation across 33 UAE visa-screening centres (PLOS One 2024; 1,309,443 radiographs, qXR v2.1) reported negative predictive values around 99.9% for normal-versus-abnormal triage at population scale. The structural caveat is scope. Almost all of this literature evaluates the comprehensive international qXR. The US FDA clearances are narrow, single-purpose devices, and the most recent of them — qXR-Detect (K251934, January 2026) — has no post-market literature yet.

Clinical Pulse

Verified clinician sentiment

Inside the Auris+ Listing

Regulatory Approvals

Safety Record

No Qure.ai qXR device recalls or FDA safety communications were identified in publicly available sources as of June 2026. The MAUDE adverse-event database could not be queried directly from this environment, so this should be read as "none found in public reporting" rather than an exhaustive audit. The principal documented risk is the same one that attends any high-sensitivity triage or detection overlay: a real false-negative floor. Published sensitivities vary by task and population (for example 0.91 for tuberculosis triage in the Lima cohort, with specificity as low as 0.32), so the absence of a flag is not a clearance and does not substitute for the radiologist's independent read. The LungIMPACT trial is a further reminder that algorithmic accuracy does not automatically translate into improved downstream outcomes when deployed as worklist prioritisation alone.

Intended Use & Indications

Procurement reviewers must keep qXR's two regulatory identities distinct, because the same brand name spans very different scopes by market. Internationally — under an EU MDR Class IIb CE mark (qXR v4.0) and an Australian ARTG listing — qXR is a comprehensive chest X-ray reporting tool that analyses adult frontal radiographs for on the order of 29 to 30 abnormality categories (the LungIMPACT trial protocol describes 29; vendor materials cite around 30), presenting findings with heatmap localisation and a structured report to support the reading clinician. It is also licensed in Canada under a Health Canada Class III medical device licence. In the United States there is no broad qXR. Instead the FDA has cleared a series of separate, narrowly-scoped 510(k) devices: qXR-BT for confirming breathing-tube position (K212690), qXR-PTX-PE for computer-aided triage of pneumothorax and pleural effusion (K230899), qXR-CTR for cardiothoracic-ratio measurement (K231149), qXR-LN for lung-nodule computer-aided detection as a second reader (K231805), and qXR-Detect, a multi-finding CADe device cleared in January 2026 (K251934). Published performance for the international reporting tool should not be read across to these US clearances without that distinction. In every jurisdiction the system is decision-support: final diagnostic and treatment decisions remain with the interpreting clinician.