Four-electrode ECG reconstruction using anatomically grounded synthetic leads: a physiological measurement framework...

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• The standard 12-lead electrocardiogram (ECG) is crucial for cardiac diagnosis but its multi-electrode requirement hinders wearable applications.
• Developing compact ECG systems is essential for continuous and ambulatory cardiac monitoring.

Purpose of the Study:

• To create an anatomically grounded framework for reconstructing a 12-lead ECG from four synthetic electrodes.
• To enable physiologically interpretable and high-fidelity ECG monitoring using fewer electrodes.

Main Methods:

• Partitioning the 12 leads into four physiological clusters based on cardiac depolarization vectors.
• Constructing synthetic electrodes using geometric vector principles and cardiac anatomy.
• Employing a hybrid convolutional neural network-Transformer architecture for waveform reconstruction.
• Training and validating the model on large ECG datasets (PTB-XL and Chapman-Shaoxing).

Main Results:

• Achieved high performance metrics: R² = 0.878, Pearson correlation ρ = 0.939, RMSE = 0.071 mV on internal data.
• External validation showed minimal performance degradation (5%).
• Preserved over 94% of waveform components and 88% of diagnostic information with high ST-segment correlation (0.964).
• Demonstrated robustness to noise (SNR ≥ 10 dB).

Conclusions:

• The proposed framework enables algorithmic feasibility for compact four-electrode ECG systems.
• The anatomically explainable approach provides a foundation for future wearable and ambulatory ECG devices.
• High diagnostic fidelity is maintained, establishing a benchmark for hardware development.