Can AI identify tuberculosis from cough audio recordings with better accuracy than human clinicians ?

Tuberculosis remains a leading infectious killer worldwide, with early diagnosis critical for treatment success. Cough sounds contain acoustic signatures unique to respiratory conditions. AI models are being developed to analyze cough recordings for specific biomarkers of tuberculosis infection. These systems could enable remote, low-cost screening in resource-limited settings. Such tools must be rigorously validated against diverse populations to ensure reliability.

Recent studies indicate that AI can identify tuberculosis from cough audio recordings with accuracy comparable to or exceeding that of trained clinicians, particularly in low-resource settings. For example, research using convolutional neural networks and transfer learning on crowdsourced cough datasets has reported sensitivities and specificities around 90–95% in detecting TB-specific acoustic biomarkers. However, these systems often rely on high-quality recordings and may struggle with confounding factors like background noise or co-occurring respiratory conditions. Deployment in real-world clinical environments remains limited, and regulatory validation is still ongoing.

— Enriched May 12, 2026 · Source: World Health Organization