Synthetic ECG signals generation: A scoping review

Area of Science:

• Biomedical Engineering
• Computational Cardiology
• Data Science

Background:

• Increasing interest in synthetic electrocardiogram (ECG) data generation within the scientific community.
• Synthetic ECG signals offer benefits for understanding cardiac electrical activity, creating unbiased datasets, and enabling data anonymization for open science.
• The need for robust methodologies to generate realistic and diverse synthetic ECG data is paramount.

Purpose of the Study:

• To conduct a scoping review of methodologies for generating synthetic ECG data.
• To analyze the limitations and possibilities of various synthetic ECG data generation techniques.
• To provide guidance for selecting the most suitable technique for specific applications.

Main Methods:

• Systematic literature search and analysis of 79 relevant studies.
• Classification of studies based on methodology (mathematical modeling, computer vision, deep generative models), number of leads, number of heartbeats, and data synthesis purpose.
• Qualitative assessment of the strengths and weaknesses of identified methodologies.

Main Results:

• Three primary categories of synthetic ECG data generation methods were identified: mathematical modeling, computer vision-based approaches, and deep generative models.
• The review highlights the diverse applications and varying levels of fidelity and variability across different synthetic ECG generation techniques.
• A significant challenge lies in establishing standardized metrics for quantitatively assessing the quality of synthetic ECG data.

Conclusions:

• The analysis provides a comprehensive overview to aid researchers in choosing appropriate synthetic ECG data generation methods.
• Further research is needed to develop standardized evaluation metrics for synthetic ECG data.
• Advancements in synthetic ECG generation are critical for accelerating research in cardiac electrophysiology and improving data sharing practices.