Project 458778

Cough is a key symptom of respiratory diseases, including tuberculosis (TB) and COVID-19. The ongoing COVID-19 pandemic has accelerated advancements in the field of digital cough monitoring using artificial intelligence (AI). Prior studies and AI models have shown that AI can identify human coughs from ambient sounds (cough detection) and can potentially differentiate coughs caused by different diseases (cough classification). For example, there is a promising smartphone application named Hyfe Research that uses AI to detect human cough, with more than 97% accuracy. Such AI models can be used on smartphones, allowing for non-invasive, easy to use tools. In this study, we will develop and evaluate a cough classification AI model which can be used on smartphones to differentiate TB and COVID-19 coughs from other types of coughs. Coughing participants presumed to have TB and/or COVID-19 will be recruited from health facilities in Lima, Peru. Clinical and demographic information will be collected from the patient, and 10 coughs will be recorded using the Hyfe Research app. Using these cough sounds, we will train algorithms to differentially identify TB and COVID-19 coughs and compare how well the algorithms perform against the laboratory reference standard. We will also conduct in-depth interviews with patients and healthcare providers to understand the feasibility and acceptability of smartphone-based cough recording in a clinical setting. The development of a reliable AI application for cough detection could improve screening strategies for respiratory infectious diseases such as TB and COVID-19, and thus mitigate future infections and outbreaks in low-resource settings.