Portable X-rays and offline AI bring TB screening into a deep gold mine

A mine is an improbable place for a screening clinic. At Mponeng in South Africa, one of the deepest gold mines in the world, workers can spend their shifts kilometres below ground, exposed to the dust and other conditions that make lung disease an occupational concern. Finding out whether a persistent cough warrants further investigation ordinarily means gaining access to equipment and clinical expertise above ground. Google Research and its partners say they have brought portable X-ray equipment and offline AI-assisted reading roughly four kilometres underground, so an initial tuberculosis screen can occur close to where people work.[1]

The deployment is part of a larger programme described by Google in a 23 October update. Google and South Africa-based Nexus AI say their tools have helped more than 90 clinicians assess over 25,000 chest radiographs across approximately 40 healthcare facilities, in a programme involving South Africa, Sierra Leone and Côte d'Ivoire and expansions under way in the Democratic Republic of the Congo, Zambia and Cameroon. They report more than 3,000 free screens for resource-constrained public clinics. These figures establish activity and access, as reported by the developers, not independently verified TB diagnoses or deaths prevented.[1]

The mine matters more than the benchmark

Portable X-ray technology is valuable only if it is at the point where people can use it. In an underground workplace, a locally functioning screening system can remove an avoidable logistical step: the worker need not first travel to a distant radiology unit simply to determine whether a chest image warrants investigation. The system is designed to process scans rapidly even without reliable connectivity. A worker whose image is concerning can then be referred for confirmatory testing. A flagged film is a reason to check, not a laboratory finding that establishes the presence of tuberculosis.[1]

Google reports image-triage results in under 45 seconds, compared with waits of weeks in some previous pathways. This is a claim about access to an initial result. It does not mean the full diagnostic pathway ends in 45 seconds: confirmatory tests, clinical decisions and follow-up remain separate steps.[1]

What the deployment numbers really measure

Regulatory clearance is not the same as WHO endorsement

Google says Nexus AI has obtained European CE certification as a Class IIb medical device and ISO 13485 certification for quality management. It has also signed an agreement with FIND to pursue WHO prequalification and is pursuing a route toward US Food and Drug Administration clearance. Those are distinct statuses. An application pathway is not a WHO prequalification certificate or an FDA clearance. The WHO independently assessed chest X-ray computer-aided detection products in 2025 and found six met its performance standards for TB screening among people aged 15 years and older; that finding must not be read as an automatic approval of every system with a TB-scoring feature.[1][2]

WHO recommends that people with a positive screen undergo confirmatory testing before TB treatment. Its 20 October guidance also calls for evidence about patient outcomes, cost, equity and feasibility alongside diagnostic accuracy. Those measures would show whether bringing screening closer to workers also shortens the route to appropriate care.[2][3]

The practical change at Mponeng is a screening service that can work where the miners are, without depending on a continuous internet connection. That removes a barrier to being checked. Following people from that first image through confirmation and treatment will establish how much further the benefit extends.[1]

Report an error or suggest a correction