Can We Triage for Pulmonary Tuberculosis from the Sound of a Cough? A Comprehensive Technical Review of Artificial Intelligence-Based Approaches

Abstract

Tuberculosis (TB) remains the world's deadliest infectious disease, and progress toward elimination is hindered by persistent gaps in early diagnosis. Current reliance on sputum-based tests excludes many vulnerable groups and constrains communitylevel detection. Recent advances in artificial intelligence (AI) and mobile health technologies have renewed interest in cough, the most universal symptom of pulmonary TB, as a scalable acoustic biomarker. Emerging studies suggest that cough carries disease-specific signatures that can be captured by digital devices and interpreted through AI models, raising the possibility of rapid, non-invasive, and widely deployable triage tools. However, the field is still constrained by small and geographically skewed datasets, inconsistent data collection methods, and limited validation in real-world populations. In this technical review, we synthesize evidence from existing studies and situate cough analysis within the broader landscape of non-sputum diagnostics. We highlight methodological and clinical challenges, examine the roles of diverse stakeholders in development and deployment, and outline a roadmap toward equitable translation. If these challenges are addressed, AI-assisted cough diagnostics could redefine TB case finding by moving testing from centralized laboratories to community settings worldwide.