A new wave of AI-driven diagnostics presented at the Union World Conference on Lung Health in Copenhagen is poised to redefine how tuberculosis (TB) is detected, monitored, and prevented worldwide. From smartphone-based cough analysis to non-invasive breath tests and child-friendly X-ray tools, these technologies target one of the world’s deadliest infectious diseases, responsible for an estimated 1.25 million deaths in 2024.
“These innovations show the extraordinary potential of AI in the fight against TB and lung disease. The challenge now is ensuring they reach the communities that need them most”, said Guy Marks, president of the International Union Against Tuberculosis and Lung Disease.
AI-powered breath analysis
Researchers from SUSTech and Shenzhen Third People’s Hospital demonstrated an AI-enabled breathomics platform capable of tracking patient recovery more sensitively than sputum cultures or X-rays. Using the AveloMask device, they collected breath samples from 60 patients in South Africa and applied machine learning to detect chemical patterns linked to treatment progress. Early detection of improvement could shorten treatment, improve adherence, and significantly reduce costs.
India’s AIIMS, JIPMER and Salcit Technologies showcased Swaasa, a smartphone-based platform that analyzes cough recordings to flag TB. Trained on more than 350 patients, the algorithm correctly identified TB-related cough signatures in 94% of cases. The low-cost and mobile design makes it ideal for remote or resource-limited settings.
Mapping vulnerability
Wadhwani Institute for AI introduced an AI-driven mapping tool that identifies communities most at risk of undiagnosed TB. By combining 20+ open datasets with anonymized surveillance data, the system accurately flagged the top 20% of high-risk villages with 71% precision, supporting more targeted outreach under India’s National TB Elimination Program.
Healthtech company Qure.ai announced regulatory clearance in Europe for qXR, the first AI chest X-ray tool approved for children from birth to 15 years. The solution supports earlier and more reliable detection in one of the hardest-to-diagnose populations: young children.
A call for rigorous validation
While optimism is high, experts stress the need for evidence-based deployment. “We need more and better tools,” said Ketho Angami of India’s ARK Foundation, “but only if accuracy and specificity are well validated. Otherwise, reliance becomes dangerous.” Proper training for health workers using AI systems is essential to ensure they interpret results safely and effectively.
Together, these innovations signal a major shift toward accessible, data-driven TB detection, bringing the promise of earlier diagnosis and more equitable care within reach for millions.
AI-ultrasound innovation
Earlier this year we reported on a new AI-powered lung ultrasound solution, ULTR-AI, that shows strong potential in accelerating tuberculosis (TB) detection, particularly in low-resource settings. Recent findings presented at ESCMID Global reveal that the deep learning suite outperforms radiologists and other human experts by 9% in identifying pulmonary TB from ultrasound images. This offers a rapid, sputum-free and scalable diagnostic alternative at a time when global TB cases have risen nearly 5% between 2020 and 2023, reversing decades of progress.
ULTR-AI is designed to work with portable, smartphone-connected ultrasound devices, enabling true point-of-care triage without reliance on expensive imaging infrastructure or specialist interpretation. According to lead researcher Dr. Véronique Suttels, the technology can be used even by minimally trained providers, making it particularly valuable in rural and understaffed regions.
