Continuous blood pressure monitoring via hemodynamic parameter and pulse transit time derived from capacitive sensing pads

Figure 1.. (a) Schematic illustrations of the capacitive sensing pads embedded with SEC sensors, sensing mechanism and equivalent circuit model. The capacitive signal is modeled as three serially connected components: CCPC (CPC electrode), Cpressure (pressure), and Cϵ (blood volume). Blood acts as a floating ground. As blood volume changes, this ground shifts, modulating capacitance. Frequency-based processing allows differentiation between blood volume changes (<0.5 Hz) and cardiac-induced vibrations (1–10 Hz). (b) Representative SEC signals from the chest and leg, with an overview of the BP estimation pipeline. Physiological features are extracted from preprocessed SEC signals, then combined with demographic information and calibration BP to train a neural network.

Figure 2.. Flowchart of the testing protocol incorporating BP-perturbing maneuvers (Valsalva and sustained isometric handgrip). Intermittent reference BP was measured during supine rest, whereas continuous reference BP was recorded throughout the testing period.

Figure 3.. Flowchart of signal processing and feature extraction for BP estimation.

Figure 4.. BP monitoring in a human study with an oscillometric arm-cuff reference (N = 30). (a) Representative results from two subjects performing Valsalva (V1, V2) and isometric handgrip (E1, E2) maneuvers, with 1 min rest periods for cuff measurements. (b) Histograms of SBP and DBP differences between capacitive pads and reference arm-cuff. Differences are normally distributed, with ∼95% within ±1.96 SD of the ME, indicating minimal subject-dependent variability. (c) Correlation and Bland–Altman analysis comparing BP estimation from the capacitive sensing pad with reference values. Pearson’s r = 0.94 for SBP and r = 0.95 for DBP (both p < 0.001). Estimation error (ME ± SD) is 0.0 ± 5.6 mmHg for SBP and 0.1 ± 3.8 mmHg for DBP; 95% limit of agreement is −11.0 to 11.0 mmHg (SBP) and −7.5 to 7.6 mmHg (DBP).

Figure 5.. BP monitoring in a human study using a finger cuff as reference (N = 8). (a) Representative BP estimation from one subject using (a) PTT + mean ΔC or (b) PTT models. BP variation maneuvers include Valsalva (V1, V2) and isometric handgrip (E1, E2). Wavelet coherence spectrograms of MBP between the pad-derived signal and finger cuff reference. The x-axis denotes time, y-axis indicates time window size, and color scale shows MSC, with higher MSC reflecting stronger coherence. Black arrows indicate phase difference (rightward: in-phase; leftward: anti-phase). White dashed lines outline statistically reliable regions.

Figure 6.. Evaluation of continuous BP monitoring in a human study using a finger cuff as reference (N = 8). (a) Comparison of average maximum MSC with the reference between bootstrapped data and pad-derived MBP. A z-score of 2.0, exceeding the 95% confidence threshold (1.96), confirms the statistical significance of the higher MSC in pad-derived data. (b) Correlation and Bland–Altman analysis comparing pad-derived BP with reference values. Pearson’s r = 0.93 for SBP and r = 0.89 for DBP (both p < 0.001). Estimation error (ME ± SD) is 0.2 ± 7.8 mmHg (SBP), 0.0 ± 4.1 mmHg (DBP); 95% limits of agreement is −15.1 to 15.6 mmHg (SBP), −8.0 to 8.0 mmHg (DBP).