Diagnostic accuracy of artificial intelligence for tuberculosis detection from cough sounds: a systematic review and meta-analysis

Tuberculosis (TB) remains the leading cause of death from infectious diseases globally, with significant diagnostic challenges in low- and middle-income countries. Artificial intelligence (AI) analysis of cough sounds could offer an inexpensive and accessible solution for detecting TB. This systematic review and meta-analysis evaluated the diagnostic accuracy of AI-based cough sound analysis for screening TB and identified key methodological gaps. We performed a systematic review (PROSPERO: CRD420250656065), searching PubMed, Scopus, IEEE, Web of Science, and CINAHL for studies published between 1 January 2009 and 31 December 2024. Included studies focused on the application of AI-algorithms for TB screening based on cough sound analysis. Risk of bias was assessed using the QUADAS-AI tool. The sensitivity, specificity, and area under the curve were extracted to quantify diagnostic performance. Overall, 14 studies were found, largely from Asia and Africa. Although a meta-analysis of seven studies showed a pooled sensitivity of 91% (95% CI: 88-94%) and a specificity of 89% (95% CI: 85-92%), with a diagnostic odds ratio of 81.61 and an area under the curve of 0.9539, indicating strong diagnostic accuracy, most of the included studies focused on analytical validity. Artificial intelligence models for cough sounds might improve TB detection, particularly in resource-limited settings, by offering a non-invasive, rapid screening tool. However, the high risk of bias, heterogeneity, and reliance on internal validation highlights the need for multicenter clinical validity studies before adoption.