Related Experiment Video
Updated: Jul 23, 2026
04:17
DNA Virus Detection System Based on RPA-CRISPR/Cas12a-SPM and Deep Learning
Published on: May 10, 2024
845
An explainable COVID-19 detection system based on human sounds.
Huining Li1, Xingyu Chen2, Xiaoye Qian3
1Department of Computer Science and Engineering, University at Buffalo, United States.
Smart Health (Amsterdam, Netherlands)
|October 24, 2022
Summary
This study introduces a machine learning approach using acoustic biomarkers from speech, cough, and breath to diagnose COVID-19. The developed system significantly improved diagnostic accuracy compared to baseline methods.
Area of Science:
- Biomedical Engineering
- Artificial Intelligence in Medicine
- Respiratory Medicine
Background:
- Acoustic signals from the human body serve as valuable biomarkers for disease diagnosis and monitoring.
- COVID-19, a respiratory illness, presents opportunities for digital acoustic biomarker investigation.
Purpose of the Study:
- To develop an accurate and explainable machine learning model for COVID-19 diagnosis using acoustic data.
- To explore the efficacy of speech, cough, and breath sounds as digital biomarkers for COVID-19 detection.
Main Methods:
- Data augmentation and Mel-spectrogram transformation were applied to acoustic datasets.
- A deep residual architecture-based machine learning model was developed for prediction.
- Activation map visualization was used for model interpretation and validation.
Main Results:
- The developed system demonstrated superior performance over baseline methods.
- A significant increase in ROC-AUC of 5.47% was achieved.
- Model interpretability was confirmed through activation map visualization.
Conclusions:
- Machine learning models utilizing acoustic biomarkers show promise for accurate and explainable COVID-19 diagnosis.
- Speech, cough, and breath analysis can serve as effective digital biomarkers for respiratory diseases like COVID-19.
