AI-Powered X-Rays Find Hidden TB Cases Among Asymptomatic People with HIV in Delhi

A person living with HIV walks into a government clinic in Delhi for a routine check-up. They feel fine. No cough, no fever, no obvious sign of trouble. A chest X-ray is taken as part of their visit—not because they complained of symptoms, but because the clinic has begun screening everyone who comes through the door. The machine, powered by artificial intelligence, reads the image in real time. What it finds changes everything.

This scene, repeated across two Delhi initiatives, has begun to reveal a hidden crisis within a crisis. Between May and August of 2026, 1,559 people living with HIV underwent X-ray screening at two government antiretroviral therapy centres. The screening identified 21 confirmed cases of active tuberculosis. Twelve of those 21 patients—57 percent—had no symptoms at all. Without the X-ray, those twelve people would have walked out of the clinic unaware they carried TB, their infection spreading silently to others around them.

The finding is not incidental. It reflects a fundamental shift in how India's TB programme now operates. Since December 2024, the government has deployed mobile vans equipped with AI-powered handheld X-rays to screen high-risk populations. In 2025 alone, this approach identified over one million asymptomatic people with active TB disease. The pattern is consistent: roughly half of all active TB cases produce no symptoms that would prompt a person to seek care. Verbal screening—asking patients if they have a cough or fever—misses them entirely.

For people living with HIV, the stakes are particularly high. Tuberculosis remains the most common opportunistic infection among them and the leading cause of death in this population. Early detection and treatment are not optional; they are survival. The Delhi initiative, led by Humana People to People India in partnership with government ART centres, embedded X-ray screening directly into routine HIV care. People attending their regular follow-up appointments could undergo chest imaging without making a separate trip. When the AI flagged a presumptive case, sputum samples were collected and sent the same day for molecular confirmation—the gold standard test that can detect TB with far greater accuracy than older microscopy methods.

The results were stark. Of 273 people identified as presumptive for TB through X-ray, all 273 underwent confirmatory molecular testing. That is 100 percent confirmation testing—a figure that stands in sharp contrast to India's national average of 38 percent in 2024. Twenty-one cases were confirmed. All were started on treatment through the government's National Tuberculosis Elimination Programme.

A second initiative reached people the health system often fails to find at all: homeless individuals, migrant workers, and residents of temporary urban settlements. Between November 2025 and August 2026, nine government chest clinics in Delhi deployed hotspot-based screening camps using the same AI-powered X-ray technology. The approach had to adapt to the population. In dispersed areas where homeless populations moved frequently, vehicle-based X-ray sites worked best. In dense slums where vehicles could not navigate, ultra-portable battery-operated machines provided the flexibility needed. Among 30,025 people screened, 4,082 were identified as presumptive for TB. Of those, 234 were confirmed to have active disease. Critically, 26,639 of the people screened had no symptoms. Of these asymptomatic individuals, 52 were confirmed to have TB.

The numbers reveal something important about how screening methods work. Verbal screening identified TB in 46.6 percent of people flagged as presumptive—a high yield because it concentrated on people already showing signs of illness. X-ray screening identified TB in only 5.7 percent of presumptive cases—a lower proportion, but a much wider net. The difference is not a failure of X-ray; it is the point. X-ray finds people who do not yet know they are sick, people whose infection remains invisible to themselves and to the health system. For a TB programme trying to interrupt transmission and prevent death, that invisible population is precisely who must be found.

The final piece was ensuring that finding TB actually led to cure. The Delhi teams paired screening with immediate access to molecular testing and treatment, backed by community health workers—the majority of them women—who provided regular follow-up and support. Among homeless and migrant populations, TB treatment completion and success rates reached 90 percent, matching the general population. The programme also incorporated TB preventive treatment, offering it to 232 close contacts of people with active disease and 1,222 vulnerable individuals at risk. The pathway was complete: identify, confirm, treat, support, prevent.

The lesson is deceptively simple but demands a fundamental reorientation of how health systems work. Disease that is invisible to the patient remains invisible to the health system unless screening methods are designed to look for it. The question is no longer only who walks into a clinic with symptoms. It is who never walks in at all, and who carries infection without knowing it. For India's TB programme, and for health systems everywhere, that distinction now shapes the difference between finding half the cases and finding them all.