Intrinsic PPG-ECG Coupling for Accurate and Low-Power Blood Pressure Monitoring.

Figures

• 

  FIGURE 1. Overall design and working principle. (a) Schematic diagram of a chest wearable sensor patch for BP monitoring, which synchronously acquires ECG and PPG signals, followed by subsequent signal processing, coupling, and wireless transmission to a terminal for continuous BP estimation. (b) Explosive view of the chest wearable sensor patch. The sensor patch is vertically stacked from the bottom (in contact with the skin) to the top in the following sequence: PPG sensor, ECG electrodes on a fabric substrate (skin‐contact interface), and a patch with an fPCB circuit and power supply. (c) Optical images of the sensor patch. (d) Block diagram illustrating the signal flow of PPG and ECG signals in the circuit of the sensor patch. The original signals are acquired and processed by the AFEs, coupled by an MCU, and wirelessly transmitted via BLE to a terminal device for further analysis and display.

• 

  FIGURE 2. Principles of signal coupling and PPG fiducial points detection. (a) Determination of the PTT, the time difference between the PPG and ECG signals. All signals were acquired with the volunteer in a seated, resting position unless otherwise stated. The ECG was recorded in a simulated Lead I configuration with electrodes on the chest, and the PPG sensor was located below the right collarbone. (b) Comparison between the PPG‐ECG coupled signal and the original ECG signal, showing that the coupled signal accurately preserves the temporal characteristics of the source signal. (c) Illustration of the coupled PPG‐ECG signal with key fiducial points (including the ECG R wave, PPG systolic peak, and onset point). (d) Identification of high‐SNR R waves via PPG‐ECG signal coupling. (e) Flowchart for extracting vital fiducial points from the coupled PPG‐ECG signal. (f) Identification of PPG fiducial points anchored to the ECG R waves. The PPG‐ECG coupled signal uses R waves of the ECG signal as reference points to locate PPG onset points (local minimum) and systolic peaks (local maximum) within each cardiac cycle.

• 

  FIGURE 3. Fiducial points detection and performance comparison. (a) Coupled PPG‐ECG signals acquired from eight volunteers. Each subplot shows F1 score improvement for systolic peak detection. (b–d) Scatter plots of PTT calculation errors against reference PTT values for (b) PPG‐ECG coupled method (n = 2560), (c) local extrema detection (n = 2560), and (d) adaptive derivative analysis (n = 182). The PPG‐ECG coupled method had significantly lower PTT errors (Kruskal‐Wallis test, *p < 0.05; post hoc Mann‐Whitney U test, *p < 0.05).

• 

  FIGURE 4. BP prediction via PPG–ECG coupling method. (a) Illustration of extracted features from the coupled PPG‐ECG signal, including the RR interval, PTT, PPG intensity ratio (PIR), and the time interval from the pulse onset point to the systolic peak (denoted as T). (b) BP estimation using an XGBoost model input with the waveform features and physical characteristics. (c) Scatter plot of SBP and DBP from 24 volunteers with a wide dynamic range, enabling robust BP estimation across diverse conditions (n = 1574). (d) Evaluation of the BP estimation model with a single volunteer's test set, comparing predicted and reference BP values over a complete measurement cycle. Scatter plots comparing predicted and reference BP estimates for systolic (e), diastolic (f), and mean arterial pressure (g) using test set data from 5 volunteers (n = 328). The RMSE and R2 were calculated for each model.

• 

  FIGURE 5. Performance evaluation of the chest wearable sensor patch with PPG–ECG coupled signal. (a) User interface of the BP monitoring system. (b) Continuous BP estimation versus reference values from a volunteer during a single test session, including an exercise interval (“Break”). (c,d) Scatter plots of estimated versus reference values for SBP and DBP, respectively, comparing the PPG‐ECG coupling method (blue) and individual signal processing (pink) (n = 221 per method). The coupling method shows significantly lower error (Wilcoxon signed‐rank test, *p < 0.05 for both SBP and DBP). (e) Box plot of ME for the coupling method. (f) Bland‐Altman and (g) violin plots representing the ME of BP measurements over a 5‐day evaluation period (n = 192).