Translation from Wearable PPG to 12-Lead ECG

TL;DR

This work tackles the lack of accessible, diagnostically rich 12-lead ECG monitoring in ambulatory settings by translating wearable PPG signals into multi-lead ECG. The authors introduce P2Es, a diffusion-based framework conditioned on demographic-aware affinity that couples a frequency-temporal forward diffusion with a multi-scale temporal reverse process and frequency deblurring. Key innovations include the GroupFinder with dynamic affinity matrices, a KNN-contrastive clustering for demographic and waveform alignment, and a lightweight diffusion scheme suitable for mobile devices, achieving state-of-the-art fidelity across time and frequency domains on MIMIC II–IV datasets. The approach demonstrates strong potential for real-time, low-cost cardiovascular surveillance and disease diagnosis outside clinical environments, preserving critical markers like ST segments and QRS morphology while enabling 12-lead ECG utility from wearables.

Abstract

The 12-lead electrocardiogram (ECG) is the gold standard for cardiovascular monitoring, offering superior diagnostic granularity and specificity compared to photoplethysmography (PPG). However, existing 12-lead ECG systems rely on cumbersome multi-electrode setups, limiting sustained monitoring in ambulatory settings, while current PPG-based methods fail to reconstruct multi-lead ECG due to the absence of inter-lead constraints and insufficient modeling of spatial-temporal dependencies across leads. To bridge this gap, we introduce P2Es, an innovative demographic-aware diffusion framework designed to generate clinically valid 12-lead ECG from PPG signals via three key innovations. Specifically, in the forward process, we introduce frequency-domain blurring followed by temporal noise interference to simulate real-world signal distortions. In the reverse process, we design a temporal multi-scale generation module followed by frequency deblurring. In particular, we leverage KNN-based clustering combined with contrastive learning to assign affinity matrices for the reverse process, enabling demographic-specific ECG translation. Extensive experimental results show that P2Es outperforms baseline models in 12-lead ECG reconstruction.

Figures (10)

• Figure 1: (a) 12-Lead ECG. (b) Components in the standard ECG signal. (c) Different ECG devices.
• Figure 2: The proposed P2Es framework.
• Figure 3: (a). Electrical axis in limb leads. (b). Principle of affinity matrix in chest leads.
• Figure 4: Information entropy analysis (TE: time-domain entropy; SE: spectral entropy).
• Figure 5: Overview of the P2Es framework: training consists of a forward process (FT DualNoise) and a reverse process (guided by PPG and the affinity matrix) for ECG reconstruction via multi-scale generation, frequency deblurring, and alignment. Inference contains only the reverse process.