Google moves TB-screening AI from benchmarks to open foundation models for point-of-care deployment

ymatias · x · 2026-10-24

At IDWeek 2026, Google Research and Google for Health detailed five years of work bringing AI to tuberculosis screening:

The problem: TB kills an estimated 1.23 million people yearly, and radiologists and imaging infrastructure are scarce in high-burden settings — from remote rural clinics in Africa to gold mines 4 km underground — leaving millions of active cases undetected.
The path: an AI tool building on Google's medical imaging work to flag presumptive TB cases for follow-up testing, prospective clinical trials, regulatory clearance with local partners, and now open foundation models like MedGemma that public health systems can tune and deploy at the point of care.

A concrete example of medical AI research maturing from benchmarks to real-world clinical deployment.

More from Apps