COVID-19 cough sound symptoms classification from scalogram image representation using deep learning models.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Signal Processing

Background:

• Deep learning models are increasingly utilized in AI for various applications.
• Visual recognition tasks often involve image ranking and artifact detection.
• Classifying respiratory illness indicators like cough sounds is crucial for public health.

Purpose of the Study:

• To develop and evaluate a deep learning model for classifying COVID-19 cough sounds.
• To differentiate between COVID-19 and healthy coughs in real-world environments.
• To assess the performance of various deep transfer learning models for cough sound analysis.

Main Methods:

• A two-step approach: sound-to-image transformation using scalogram and feature extraction/classification.
• Utilized six deep transfer learning models: GoogleNet, ResNet18, ResNet50, ResNet101, MobileNetv2, and NasNetmobile.
• Trained and tested on a dataset of 1457 cough sounds (755 COVID-19, 702 healthy).

Main Results:

• The proposed model achieved a maximum accuracy of 94.9% with the SGDM optimizer.
• ResNet18 emerged as the most stable model for classifying cough sounds from a limited dataset.
• ResNet18 achieved a sensitivity of 94.44% and a specificity of 95.37%.

Conclusions:

• The developed deep learning model shows significant promise for reliable and accurate cough sound classification.
• The ResNet18 model is particularly effective for analyzing limited cough sound datasets.
• Further research with larger datasets is warranted to test the model's generalization capabilities.