Acoustic Artificial Intelligence for Tuberculosis Screening: Machine Learning Analysis of Cough Sounds in Resource-constrained environments

TB-ACOUSTIC Africa is a machine learning–based tuberculosis screening system that uses cough and respiratory sound analysis to support non-sputum tuberculosis screening in low-resource settings. The system processes audio recordings, extracts acoustic features such as Mel-frequency cepstral coefficients (MFCC), spectral features, zero-crossing rate, and temporal features, and uses machine learning models to detect cough events and classify respiratory sounds. The system is designed as a low-cost, deployable screening and decision-support tool that can be integrated into mobile health applications and community health screening programs to support early tuberculosis screening and referral in settings with limited laboratory and radiographic infrastructure.