Respiratory Diseases Diagnosis Using Audio Analysis and Artificial Intelligence: A Systematic Review

Area of Science:

• Medical Informatics
• Bioacoustics
• Machine Learning

Background:

• Respiratory diseases pose a significant global health challenge, requiring advanced diagnostic tools.
• Audio-based digital biomarkers from respiratory sounds and voice are emerging as key indicators of respiratory health.
• Machine learning (ML) provides powerful methods for analyzing these complex audio signals.

Approach:

• This review synthesizes findings from 75 studies focused on audio analysis for respiratory conditions.
• The analysis categorizes research into cough detection, lower respiratory symptom identification, and voice/speech diagnostics.
• Publicly available datasets relevant to respiratory audio analysis are also presented.

Key Points:

• ML algorithms are increasingly used to extract diagnostic information from respiratory sounds and voice.
• Studies address challenges like cough sound recognition in noisy environments and detecting wheezes or crackles.
• Voice and speech analysis are explored for evaluating voice abnormalities related to respiratory issues.

Conclusions:

• Audio-based digital biomarkers and ML show significant potential for non-invasive respiratory disease diagnosis.
• Research trends are influenced by the COVID-19 pandemic, accelerating studies in remote diagnostics and mobile data acquisition.
• Further development in this field promises improved, accessible respiratory care.