PG-LRF: AI Model Generates ECG from Wearable PPG Signals
A novel framework named PG-LRF has been introduced by researchers, designed to generate electrocardiogram (ECG) signals from photoplethysmography (PPG) data, guided by physiological principles. While ECG is the gold standard for heart evaluation, it necessitates specialized equipment, unlike PPG, which is commonly found in wearable devices but suffers from noise and lacks clear diagnostic features. Current methods for converting PPG to ECG depend on statistical alignment and do not incorporate physiological constraints. PG-LRF employs an electro-hemodynamic simulator that simultaneously models ECG and PPG using shared cardiac phase dynamics, directed by a Physiology-Aware AutoEncoder. This approach enhances generation accuracy by structuring the latent space around physiology-informed electro-hemodynamic factors. The study is available on arXiv with the identifier 2605.12541.
Key facts
- PG-LRF stands for Physiology-Guided Latent Rectified Flow.
- It generates ECG from PPG signals.
- ECG is the clinical standard for cardiac assessment.
- PPG is ubiquitous in wearables but lacks ECG-specific morphology.
- Existing methods fail to explicitly structure latent space around physiology-aware factors.
- PG-LRF introduces an electro-hemodynamic simulator.
- The simulator co-models ECG and PPG through shared cardiac phase dynamics.
- The paper is on arXiv with ID 2605.12541.
Entities
Institutions
- arXiv