[Opening scene: a community health worker at a rural clinic.]

Problem

Tuberculosis kills more people than any other single infection. Many first-contact clinics have no laboratory and no X-ray machine, so TB is often caught late.

AI solution

Community health workers with little training scan the lungs with a portable ultrasound. An AI model that runs offline reads the images and flags likely TB. The whole system costs under $5,000, according to the team.

Impact

The team reports accuracy that meets the World Health Organization’s targets for a TB triage test. A 60-month trial with 3,000 adults across Benin, Mali and South Africa is planned. [NEED: published validation results and the trial’s status in 2026.]

People

[The LiGHT lab team and the health workers testing it.]

This story supports

SDG3

Sources (1)

ITU, AI for Good Innovate for Impact Interim Report 2026, pp. 1320–1326

Spot an error? Tell us.