A Physics-Aware Diffusion Framework for Robust ECG Synthesis Using Mesoscopic Lattice Boltzmann Constraints.

Area of Science:

• Biomedical Engineering
• Computational Physiology
• Medical Signal Processing

Background:

• Cardiovascular disease is a leading global cause of death, necessitating advanced cardiac monitoring solutions.
• Electrocardiogram (ECG) is the gold standard but complex to acquire long-term; Photoplethysmography (PPG) from wearables is accessible but synthesizing ECG from it is challenging.
• Current data-driven methods for ECG synthesis from PPG lack physical constraints, leading to non-physiological outputs.

Purpose of the Study:

• To develop a novel physics-aware framework for accurate and interpretable ECG synthesis from PPG signals.
• To address the limitations of purely data-driven approaches by incorporating biophysical mechanisms.
• To improve the feasibility of long-term cardiac monitoring using wearable PPG devices.

Main Methods:

• Proposed PhysDiff-LBM, a conditional diffusion model integrating Lattice Boltzmann Method (LBM) hemodynamic constraints.
• Utilized a dual-stream architecture with cross-attention for high-frequency details and region-wise adaptability.
• Physically regularized ECG synthesis by enforcing consistency with hemodynamic momentum and fluid dynamic conservation laws.

Main Results:

• Achieved superior signal fidelity in synthesized ECG waveforms compared to existing methods.
• Demonstrated significant advantages in downstream clinical applications.
• The physics-aware approach reduced non-physiological artifacts and improved interpretability.

Conclusions:

• PhysDiff-LBM offers a robust and physically grounded method for ECG synthesis from PPG.
• The framework enhances the reliability of cardiac monitoring, particularly for long-term applications using wearable technology.
• Integrating hemodynamic constraints improves the physiological plausibility and clinical utility of PPG-derived ECG signals.