Tuberculosis kills more people worldwide than any other single infectious agent. In 2023, the World Health Organization reported 1.25 million TB deaths globally, a number that has barely budged in a decade. South Africa bears one of the heaviest burdens, ranking among the top ten countries for TB incidence. In rural areas, where clinics are distant and radiologists are scarce, many cases go undiagnosed until the disease has advanced and become highly contagious.
A Mumbai-based company called Qure.ai is changing that equation. Its product, qXR, is a deep learning system that reads chest X-rays for signs of tuberculosis. In a deployment across South African mobile screening vans, qXR screened 6,500 people and identified 187 cases of TB that would likely have been missed by the existing health system. Each X-ray is processed in under one minute.
The Missing Millions
The WHO estimates that roughly 3 million people with TB go undiagnosed every year. These "missing millions" continue to spread the disease in their communities, each untreated person potentially infecting 10 to 15 others annually. The diagnostic gap exists not because TB is hard to treat (a standard course of antibiotics cures most cases) but because it is hard to find.
In South Africa, the traditional screening pathway involves a symptom questionnaire followed by sputum collection for laboratory testing. The problem is that symptom-based screening misses up to half of active TB cases, particularly in people co-infected with HIV, where cough and weight loss may be attributed to other causes. Laboratory testing is accurate but slow. It can take days or weeks for results to come back, and many patients never return to collect them.
X-rays Without Radiologists
Chest X-rays are one of the most effective tools for TB screening. They can detect lung abnormalities consistent with TB even in patients who are not yet showing symptoms. But reading X-rays requires trained radiologists, and South Africa has approximately one radiologist per 100,000 people. In rural provinces like Limpopo and the Eastern Cape, the ratio is far worse.
Qure.ai's qXR system fills this gap. The software runs on a standard laptop connected to a portable X-ray machine. A patient walks in, gets an X-ray, and within seconds the AI flags whether the image shows signs consistent with TB. Flagged patients are immediately referred for confirmatory testing with GeneXpert, a rapid molecular diagnostic that can confirm TB and detect drug resistance in under two hours.
The AI does not replace the confirmatory test. What it does is function as a highly sensitive triage tool, deciding who needs further testing and who can be cleared. In clinical studies, qXR has demonstrated sensitivity above 95%, meaning it catches the vast majority of true TB cases while maintaining a specificity high enough to avoid overwhelming the confirmatory testing pipeline.
Mobile Clinics Reach the Unreachable
The South Africa deployment uses mobile screening vans that travel to communities, mine sites, and informal settlements. The vans are equipped with digital X-ray machines, laptops running qXR, and GeneXpert devices. A team of community health workers recruits participants, conducts the screening, and provides same-day results for those who test positive.
In the screening of 6,500 people, 187 active TB cases were identified. Many of these individuals had no symptoms or had symptoms they had not associated with TB. Without the mobile vans and AI screening, they would have continued in their communities undiagnosed, spreading the disease to family members, coworkers, and neighbors.
Scaling Across the Continent
Qure.ai has now deployed qXR in more than 30 countries. The WHO issued a recommendation in 2023 endorsing the use of computer-aided detection (CAD) for TB screening, specifically citing qXR and similar products as appropriate tools for community-based active case finding. This recommendation opened the door for national TB programs to include AI screening in their standard protocols and for global health funders like the Global Fund and PEPFAR to support deployment at scale.
The economics are compelling. A chest X-ray with AI interpretation costs a fraction of what it would cost to employ a radiologist for the same task. The portable equipment fits in a van. The software works offline with periodic updates. For countries where TB kills tens of thousands annually and health budgets are stretched thin, the combination of mobile X-ray and AI triage offers a path to finding and treating cases that the health system currently misses.
Every Missed Case Has a Cost
Each undiagnosed TB case represents not only suffering for the individual but a chain of transmission that sustains the epidemic. Modeling studies suggest that finding and treating one infectious case of TB prevents an average of five secondary infections. The 187 cases found in this single South African deployment, if treated successfully, could prevent nearly a thousand future infections.
A van. A portable X-ray machine. A laptop running an AI model trained on millions of chest images. This is what closing the diagnostic gap looks like in practice. Not a theoretical breakthrough in a research lab, but a community health worker driving down a dirt road to find the patients the system forgot.
Sources: Qure.ai qXR clinical validation studies; WHO operational handbook on tuberculosis, Module 2: Screening (2023 update); WHO Global Tuberculosis Report 2024; South Africa National TB Programme reports; The Lancet Digital Health, systematic reviews of CAD for TB screening.
