Development of Android-Based Pulmonary Monitoring System for Automated Lung Auscultation Using Long Short-Term Memory (LSTM) Network with Post-Processing from Edge Impulse

Abstract

Chronic pulmonary diseases remain a prevalent threat globally. With the emergence of COVID-19 and its transmission, there has been a rapid increase in the number of deaths due to respiratory illnesses. In this study, lung sound classifications were performed using a Thinklabs One digital stethoscope and through the utilization of Long Short-Term Memory (LSTM) in the classification of a person's lung auscultation record into either the normal, crackle, wheeze, or stridor categories with a 92.50% accuracy. Performance evaluation of this system was also done to cross-check for the validity of the algorithm modeled through Edge Impulse, which provided a 92.77% accuracy. The integration of the system adopted an Android-based mobile application as the pulmonary monitoring platform that records a person's general respiratory health data. The inputs from the mobile application were anonymously stored in a centralized database system correspondingly for post-processing and analysis.