Wrist ballistocardiography and invasively recorded blood pressure in healthy volunteers during reclining bike exercise

Figures

• 

  FIGURE 1. Left: Setup during semirecumbent sampling. (A) The investigational device is visible on the participant’s right wrist. (B) The pressure transducer is fixed to the upper non-dominant arm, which is rested on the handle of the recumbent bike. Photo by written consent. Right: the investigational device. (C) Illustration of the tonometer wristband. (D) Sensor placement on the dominant arm.

• 

  FIGURE 2. Graphical overview of data collection and data processing pipeline. (A) Schematic representation of exercise protocol. Each participant exercised freely at four different load levels from 0 to 150 Watts as defined by work delivered to the reclining ergometer bike, set via programmable dynamic resistance at 60 RPM. The ramping protocol was repeated for supine and semirecumbent posture with a rest and recovery period between. Rest took place in the semirecumbent posture. (B) Plot of ABP for one participant over the full session, roughly vertically aligned with the respective load states. (C) Raw accelerometer data is segmented into individual beats using systolic peaks as reference. Several segments form an ensemble average. (D) The average signal is decomposed via EEMD. (E) Time-frequency representation of the ensemble average and one of the IMFs. Top: The original signal. Bottom: Processed signal, emphasizing energy in the 10–30 Hz band. A temporally localized energetic feature appears to be isolated around the 200 ms mark. (F) Ensemble averaged signals between various exercise conditions, showing differences in wave morphology. Plotting the filtered BCG average against the…

• 

  FIGURE 3. Ensemble averages from a male participant across all 8 exercise conditions. Top row is supine; bottom row is semirecumbent. At top, the averaged left ventricle outflow tract (LVOT) flow velocity shows onset of rapid ejection around 0.15–0.2 s followed by the “click” of the aortic valve closing at approximately 0.3–0.45 s. Below, the raw ensemble average of the accelerometer, followed by the filtered accelerometer signal with RMS envelope of the signal energy. Averaged arterial blood pressure (ABP) in the radial artery is shown together with the average tonometer signal, scaled to the magnitude of the ABP signal. Bottom: analytic morlet scalogram of the ensemble averaged accelerometer signals between 10 and 30 Hz. The sharp feature in the scalograms between 0.2 and 0.3 s seems to be clearly represented in the filtered BCG and is tentatively associated with rapid ejection from the ventricle.

• 

  FIGURE 4. Hemodynamics between exercise states. (A) box charts showing distribution of systolic blood pressure (SBP), diastolic blood pressure (DBP), mean arterial pressure (MAP), heart rate (HR), BCG-PPG pulse transit time (PTT) and systolic BCG amplitude. Circles are outliers. Stars indicate significance of difference between postures via two-tailed Wilcoxon signed rank test at two different significance levels: **: a < 0.01, *: a < 0.05. (B) Top: scatterplot of SBP vs. BCG-PPG PTT with least-square regression lines. Bottom: scatterplot of BCG-PPG PTT vs HR. ρ is Pearson’s correlation coefficient. (C) Top: bar chart of normalized left ventricle stroke volumes versus BCG amplitude across exercise conditions. Bottom: scatterplot of systolic BCG amplitude versus left ventricle stroke volume (SV). SV data was only available from a subset of participants. OL indicates outliers determined via the robust Mahalanobis distance at a chi-square significance level of 0.05. SV and amplitude data were normalized by z-scoring across the entire sample.