WAVE: Wall-Aligned Vector Embedding for Self-Supervised Learning of Electrocardiograms

Area of Science:

• Artificial Intelligence
• Biomedical Engineering
• Cardiology

Background:

• Deep learning significantly advances electrocardiogram (ECG) analysis.
• High annotation costs due to reliance on labeled data limit deep learning applications in ECG.
• Developing label-efficient methods is crucial for widespread ECG analysis.

Purpose of the Study:

• To propose wall-aligned vector embedding (WAVE), a self-supervised learning framework for ECG analysis.
• To reduce the dependence on labeled ECG data through effective prior knowledge extraction.
• To enhance the efficiency and reduce the cost of cardiovascular diagnostics.

Main Methods:

• WAVE framework leverages multi-lead ECGs, considering lead diversity, synergy, and correlation with cardiac walls.
• A multi-branch network captures lead-wise diversity; shared projection models wall-wise synergy.
• A dual alignment task learns correlations within and across cardiac walls.

Main Results:

• WAVE consistently outperforms existing baselines across various evaluation settings.
• The framework demonstrates strong performance with limited labeled ECG data.
• Key components like dual alignment and shared projection significantly improve pretraining quality.

Conclusions:

• WAVE effectively extracts informative ECG representations from unlabeled data.
• The proposed method enables low-cost and label-efficient ECG analysis.
• WAVE holds significant potential for real-world cardiovascular diagnostics.