State-of-the-Art Deep Learning Methods on Electrocardiogram Data: Systematic Review

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Cardiology

Background:

• Electrocardiogram (ECG) is a key noninvasive diagnostic tool.
• Deep learning (DL) is advancing diagnostic model development using physiological signals.

Purpose of the Study:

• To systematically review DL methods applied to ECG data.
• To cover diverse clinical applications of DL in ECG analysis.

Main Methods:

• Systematic PubMed search using "deep learning" and "ECG" related keywords.
• Exclusion of irrelevant articles, non-English manuscripts, and studies lacking quantitative evaluation.
• Screening of titles and abstracts, followed by full-text review.

Main Results:

• 230 relevant articles (2020-2021) identified across 6 applications: blood pressure estimation, cardiovascular disease diagnosis, ECG analysis, biometric recognition, sleep analysis, and other clinical analyses.
• Overview of state-of-the-art DL strategies and major ECG data sources.
• Identified research gaps, including addressing blood pressure variability and DL model design.

Conclusions:

• This review offers insights into current DL methods for ECG data.
• Highlights future research directions for robust DL models in clinical decision support.