Introduction:
Coughing is a common symptom associated with various illnesses, including COVID-19. Researchers are investigating the potential of using cough sound signals for a cost-effective method of disease diagnosis. Unlike traditional diagnostic methods, which are often expensive and require specialized personnel, smartphone-based cough analysis offers a more accessible alternative.
Objective:
Methods:
| Cough Sample Collection | Acoustic Analysis | Machine Learning |
|---|---|---|
| • Analyzed cough sounds from 1183 COVID-19-positive patients. • Compared with 341 non-COVID-19 cough samples. Distinguished between coughs associated with pneumonia and asthma. |
• Optimized frequency ranges to identify specific frequency bands. • Conducted statistical separability tests to validate findings |
• Used linear discriminant analysis and k- nearest neighbors classifiers. • Confirmed distinct frequency bands in cough signal power spectra associated with different respiratory diseases. |
Results:
Frequency Bands Identification
Specific frequency bands were correlated with each respiratory ailment.
Validation
Statistical tests and machine learning algorithms confirmed the presence of unique acoustic signatures for each disease.
Conclusion:
Acoustic signatures in cough sounds could revolutionize the classification and diagnosis of respiratory diseases.
This method offers a cost-effective and widely available healthcare tool for disease detection.
To know more: https://doi.org/10.1038/s41598-023-50371-2
Reference: Ghrabli S, et al. Identifying unique spectral fingerprints in cough sounds for diagnosing respiratory ailments. Scientific Reports. 2024 Jan 5;14(1):593.