Google Research detailed its ongoing development of health artificial intelligence tools and open foundation models for tuberculosis screening at IDWeek 2026 in Washington, D.C. The initiatives combine open-source vision-language models with certified diagnostic partnerships to accelerate frontline detection.
Clinical Partnerships and Regulatory Milestones
Google Research reported that Pretoria-based medical technology company Nexus AI translated its chest X-ray algorithms into a certified clinical diagnostic tool. Nexus AI secured CE MDR Class IIb certification under European Union regulations, as well as compliance certification with ISO 13485 for medical device quality management.
The technology is currently deployed across approximately 40 healthcare facilities in six countries, including South Africa, Sierra Leone, and Côte d'Ivoire. According to the announcement, the system has evaluated more than 25,000 chest radiographs, reducing turnaround times to under 45 seconds. Deployments include an ultra-portable digital X-ray setup at the Mponeng Gold Mine in South Africa.
Open-Source Foundations and MedGemma
To scale impact beyond individual partnerships, Google Research introduced the Health AI Developer Foundations, open-sourcing chest X-ray embeddings and MedSigLIP. MedSigLIP is a lightweight vision-language model aligned to match medical images with clinical descriptions.
In addition, Google released MedGemma, a family of open-weight multimodal models in 4B and 27B parameter configurations designed for local, on-device clinical reasoning. The company stated that MedGemma has achieved over 10 million downloads worldwide. In Southeast Asia, the Ministry of Health of Indonesia is utilizing MedGemma and MedSigLIP to support its national tuberculosis screening initiative.
Acoustic Screening with HeAR
Google Research also developed Health Acoustic Representations to enable non-invasive acoustic screening using cough sounds recorded on standard smartphones. HeAR is being evaluated across sub-Saharan Africa through collaborative research with organizations including CIDRZ, the Aurum Institute, and HELINA.
Key facts and where they come from
- Google Research presented its health AI developments for TB screening at IDWeek 2026 in Washington, D.C.
This week at IDWeek 2026 in Washington, D.C., we are sharing how we’re bringing health AI research to reality
- Nexus AI achieved CE MDR Class IIb certification for its diagnostic software based on Google's algorithms.
Nexus AI achieved Class IIb certification — a rigorous designation that verifies the tool is safe for real-world patient care
- Google open-sourced chest X-ray embeddings and MedSigLIP through Health AI Developer Foundations.
Through the Health AI Developer Foundations (HAI-DEF), we open-sourced our chest X-ray embeddings and MedSigLIP
- The MedGemma family of open-weight multimodal models has exceeded 10 million downloads worldwide.
Since its launch, MedGemma has seen more than 10 million downloads worldwide.
- Indonesia's Ministry of Health is fine-tuning MedGemma and MedSigLIP for its national tuberculosis screening program.
the Ministry of Health of Indonesia is addressing the world's second-highest TB burden with an ambitious goal: screening 50 million citizens annually
