Performance evaluation of the smartphone-based AI cough monitoring app - Hyfe Cough Tracker against solicited respiratory sounds

Background: Emerging technologies to remotely monitor patients’ cough show promise for various clinical applications. Currently available detection systems all represent a trade-off between convenience and performance. The accuracy of such is highly contingent on the settings in which they are intended be used. Moreover, establishing gold standards measure this challenging. Objectives: We present first performance evaluation study Hyfe Cough Tracker app, passive monitoring smartphone application. evaluate using continuous audio recordings counting by trained individuals as standard. propose standard procedures use multi-observer sound annotation from evaluating automated devices. Methods: This was embedded larger digital acoustic surveillance (clinicaltrial.gov NCT04762693). Forty-nine participants were included instructed produce diverse series solicited sounds 10-minute sessions. Simultaneously, recording performed MP3 recorder two smartphones running app monitored identified events. All independently labeled three medically-trained researchers. Results: showed sensitivity 91% specificity 98% with very high correlation rate measured that human annotators (Pearson 0.968). A standardized approach establish an identifying multiple observers presented. Conclusion: new smartphone-based system. can detect, record count coughs cough-like explosive controlled environments accuracy. Additional steps required validate system community settings.