Evaluating and Visualizing the Contribution of ECG Characteristic Waveforms for PPG-Based Blood Pressure Estimation

Area of Science:

• Biomedical Engineering
• Cardiovascular Technology
• Signal Processing

Background:

• Non-invasive continuous blood pressure (BP) monitoring is crucial for managing cardiovascular diseases (CVDs).
• Photoplethysmogram (PPG) and electrocardiogram (ECG) signals are used for BP estimation, but ECG's waveform features are underutilized.
• Current models primarily focus on PPG signal features, with ECG's R-wave peak serving mainly as a temporal reference.

Purpose of the Study:

• To evaluate the impact of specific ECG waveforms (P wave, QRS complex, T wave) on enhancing BP estimation accuracy.
• To compare the contribution of different ECG waveform components when integrated with PPG signals.
• To investigate the attention distribution of convolutional neural networks (CNNs) on PPG and ECG waveforms during BP estimation.

Main Methods:

• Utilized PPG as the primary signal, forming five input combinations by adding ECG, P wave, QRS complex, T wave, or none.
• Employed five common CNN models to assess the consistency of waveform contributions.
• Applied Gradient-weighted class activation mapping (Grad-CAM) for visualizing CNN attention to signal waveforms.

Main Results:

• CNNs demonstrated greater attention towards the QRS complex and T wave in ECG signals.
• The QRS complex and T wave showed a more significant contribution to error reduction compared to the P wave.
• Adding QRS complex and T wave individually reduced mean absolute error (MAE) for systolic BP (SBP) to 6.57 mmHg and 6.21 mmHg, and for diastolic BP (DBP) to 3.65 mmHg and 3.73 mmHg, respectively.

Conclusions:

• The QRS complex and T wave are critical ECG components for improving continuous BP estimation.
• Feature extraction from ECG's QRS complex and T wave warrants further investigation, similar to PPG waveform analysis.
• Integrating detailed ECG waveform analysis alongside PPG offers a promising avenue for more accurate non-invasive BP monitoring.