Cough monitoring and pneumonia diagnosis algorithm through analysis of respiratory system-based vibro-acoustic signals and AI technology

Abstract

In case of pneumonia often accompanied by serious complications, sometimes lead to death, early diagnosis and continuous monitoring can greatly reduce the dangerousness. Moreover, the COVID-19 pandemic has demonstrated the need for new diagnostic tools that can minimize medical personnel engagement while avoiding equipment being exposed to afflicted patients. In this study, we developed cough monitoring algorithm by detecting the vibrations of human body. The acceleration response at each part of body was measured to determine propagation characteristics of vibration when cough occurs. And it was confirmed that the monitoring accuracy was improved when use the vibration signal compared to the case of using only acoustic signal. After that, we analyzed the perceived cough in terms of psych-acoustical and sound-energy aspects. For the characteristic features derived by quantifying the results of analysis, the data augmentation process was applied, and finally AI-based pneumonia diagnosis algorithm was constructed. To estimate the performance of algorithm, the accuracy of pneumonia determination in new cough cases was verified. It showed the higher value than the accuracy of pulmonologists with only cough sounds. Therefore, developed algorithm that perform continuous cough monitoring and reliable pneumonia diagnosis can be used as an effective supplementary tool for early diagnosis and prognosis of pneumonia.