COVID-19 Detection Using Photoplethysmography and Neural Networks.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiovascular Physiology

Background:

• Early identification of microvascular changes in COVID-19 patients is crucial for clinical management.
• Photoplethysmography (PPG) signals offer a potential non-invasive window into microcirculation.
• Deep learning presents opportunities for analyzing complex physiological signals.

Purpose of the Study:

• To develop a deep learning-based method for identifying COVID-19 patients using raw PPG signals.
• To assess the feasibility of using pulse oximeter data for COVID-19 detection.
• To investigate PPG as a tool for early recognition of SARS-CoV-2-induced microvascular alterations.

Main Methods:

• Acquisition of PPG signals from 93 COVID-19 patients and 90 healthy controls using finger pulse oximeters.
• Development of a template-matching algorithm to filter noise and motion artifacts from PPG signals.
• Training a custom convolutional neural network (CNN) model for binary classification of COVID-19 versus control samples.

Main Results:

• The CNN model achieved 83.86% accuracy and 84.30% sensitivity in identifying COVID-19 patients on test data.
• A template-matching method effectively selected high-quality PPG signal segments for analysis.
• The model demonstrated robust performance in distinguishing between COVID-19 and healthy control groups.

Conclusions:

• Deep learning analysis of PPG signals can effectively identify COVID-19 patients.
• Photoplethysmography is a promising non-invasive tool for microcirculation assessment and early detection of COVID-19-related vascular changes.
• This low-cost, user-friendly approach is suitable for resource-limited healthcare settings.