Using Machine Learning Algorithms to Determine the Post-COVID State of a Person by Their Rhythmogram

Area of Science:

• Cardiology
• Medical Technology
• Infectious Diseases

Background:

• Post-COVID conditions affect cardiac function, necessitating reliable detection methods.
• Electrocardiogram (ECG) data offers a non-invasive window into heart rhythm regulation.
• Current diagnostic tools may not fully capture the nuances of post-COVID cardiac changes.

Purpose of the Study:

• To introduce a novel method for detecting post-COVID conditions using ECG.
• To identify specific ECG markers, termed "cardiospikes," indicative of a past COVID-19 infection.
• To explore the potential of these cardiospikes as objective markers for COVID-specific heart rhythm regulation.

Main Methods:

• Utilized a convolutional neural network (CNN) to analyze ECG data.
• Developed a detection algorithm for identifying "cardiospikes" in post-COVID patients.
• Conducted blood parameter measurements and created profiles for recovered COVID-19 patients.

Main Results:

• Achieved 87% accuracy in detecting cardiospikes within a test sample.
• Confirmed that cardiospikes are inherent physiological signals, not hardware artifacts.
• Established correlations between ECG findings and blood parameter profiles in recovered patients.

Conclusions:

• The novel CNN-based method effectively detects cardiospikes, indicating post-COVID cardiac changes.
• Cardiospikes show promise as reliable biomarkers for COVID-19's impact on heart rhythm.
• Findings support the development of remote screening tools for COVID-19 diagnosis and monitoring.