An evolutionary ensemble learning for diagnosing COVID-19 via cough signals

Area of Science:

• Medical Diagnostics
• Signal Processing
• Artificial Intelligence

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Cough sounds are a potential biomarker for respiratory diseases, including COVID-19.

Purpose of the Study:

• To develop and evaluate an algorithm for diagnosing COVID-19 using cough signals.
• To leverage ensemble methods and deep learning for enhanced cough signal analysis.

Main Methods:

• An ensemble algorithm combining statistical and convolutional neural network (CNN) feature extractors.
• Application of Fourier, wavelet, and Hilbert-Huang transforms for signal processing.
• Optimization of base-learner weights using an evolutionary paradigm and memetic algorithms for CNN training.

Main Results:

• The proposed algorithm demonstrated superior performance in diagnosing COVID-19 from cough signals compared to existing methods.
• Experiments included comparisons with various CNN architectures and generic machine learning algorithms.

Conclusions:

• COVID-19 can be effectively diagnosed through the analysis of cough signals.
• Convolutional Neural Networks show promise in processing cough signals for disease detection, with potential for further improvement through architecture optimization.