A collaborative effort between a multidisciplinary team from the University of Washington in Seattle and researchers at the Respiratory Disease Research Center in Nairobi has produced a mobile diagnostic tool to help detect tuberculosis. Publicly reported in Science Advances, the project describes a smartphone software solution that can assist in identifying tuberculosis through audio analysis rather than relying solely on expensive clinical equipment.
The program, called TBscreen, uses machine learning to interpret patterns in cough sounds. It relies on a trained model that processes audio data captured with a standard smartphone microphone, transforming raw sound into indicators that may signal a tuberculosis infection. This approach demonstrates how everyday devices can be repurposed to support health screening, especially in places with limited access to traditional diagnostic infrastructure.
The validation study behind TBscreen examined a substantial set of cough recordings from two groups: patients with active pulmonary tuberculosis and individuals with other respiratory conditions. Thousands of cough samples were analyzed, including both naturally occurring and elicited coughs, across a diverse patient cohort. The researchers evaluated the tool’s ability to distinguish active tuberculosis from non tuberculosis coughs and reported an overall accuracy that suggests the technology can function as a useful screening aid in real-world settings. They also found that the smartphone-based system, paired with a standard microphone, performed on par with more specialized hardware, highlighting the potential for scalable, low-cost screening in clinics and community settings across North America and beyond.
Global tuberculosis statistics continue to stress the importance of broadening diagnostic reach. The World Health Organization has pointed to millions of new infections and significant mortality, underscoring the need for innovative strategies to identify cases earlier and connect people to effective treatment. Innovations like TBscreen fit into broader public health goals by offering a fast, noninvasive screening option that can be used at the point of care or in remote communities. This complements traditional laboratory methods and helps close gaps in detection and care pathways in both Canada and the United States.
Beyond TB alone, the TBscreen story adds to growing evidence about how digital health tools can augment conventional medicine. As researchers and clinicians explore the strengths and limitations of audio-based diagnostics, questions remain about how to integrate such tools into standard practice, protect data privacy, and maintain accuracy across diverse populations and environments. The momentum behind this line of inquiry reflects a wider move toward making health screening more accessible through everyday devices, while staying mindful of the need for rigorous validation, clear clinical guidelines, and ongoing monitoring of performance in day-to-day use.