Cross-domain Joint Dictionary Learning for ECG Inference from PPG

TL;DR

This paper introduces cross-domain joint dictionary learning frameworks to improve ECG inference from PPG signals, enabling better cardiac monitoring and diagnosis through enhanced signal representation and disease information integration.

Contribution

The paper proposes novel dictionary learning models, XDJDL and LC-XDJDL, that enhance ECG inference accuracy from PPG by optimizing paired signal dictionaries and incorporating disease labels.

Findings

Outperforms previous methods in ECG inference accuracy.

Achieves better visual and quantitative results on diverse ECG/PPG data.

Demonstrates potential for PPG-based ECG screening.

Abstract

The inverse problem of inferring electrocardiogram (ECG) from photoplethysmogram (PPG) is an emerging research direction that combines the easy measurability of PPG and the rich clinical knowledge of ECG for long-term continuous cardiac monitoring. The prior art for reconstruction using a universal basis has limited fidelity for uncommon ECG waveform shapes due to the lack of rich representative power. In this paper, we design two dictionary learning frameworks, the cross-domain joint dictionary learning (XDJDL) and the label-consistent XDJDL (LC-XDJDL), to further improve the ECG inference quality and enrich the PPG-based diagnosis knowledge. Building on the K-SVD technique, our proposed joint dictionary learning frameworks aim to maximize the expressive power by optimizing simultaneously a pair of signal dictionaries for PPG and ECG with the transforms to relate their sparse codes and disease information. The proposed models are evaluated with 34,000+ ECG/PPG cycle pairs containing a variety of ECG morphologies and cardiovascular diseases. We demonstrate both visually and quantitatively that our proposed frameworks can achieve better inference performance than previous methods, suggesting an encouraging potential for ECG screening using PPG based on the proactive learned PPG-ECG relationship.