Blood Pressure Continuous Measurement through a Wearable Device: Development and Validation of a Cuffless Method

Figures

• 

  Figure 1. Worn reference and test devices. The reference device inflatable cuff is worn on the left arm and connected with an air hose to the measurement instrument placed in a cross-body bag. The sensorized shirt is worn, thanks to a zipped opening on the side, and sensors adherence is assured with elastic bands. The prototypal wristband is worn on the left wrist, in correspondence of the radial artery and connected to the sensorized shirt with a dedicated connector.

• 

  Figure 2. Extracted features from acquired ECG and PPG physiological signals: (a) PPG time-delay features; (b) PPG morphological features.

• 

  Figure 3. PPG features extraction algorithm flowchart.

• 

  Figure 4. The 30 s window extraction logic: (a) previous logic; (b) following logic; (c) previous and following logic.

• 

  Figure 5. On the top, 2 min window of filtered ECG and PPG signals around the reference device measurement. On the bottom, a zoom on the 30 s window extracted according to the previous logic.

• 

  Figure 6. PPG features extraction logic from each 30 s considered window.

• 

  Figure 7. Variation of Systolic Blood Pressure (ΔSBP), Diastolic Blood Pressure (ΔDBP), Mean Blood Pressure (ΔMBP) and Pulse Arrival Time (ΔPAT) in the different protocol phases: ΔEffort, ΔRecovery and ΔRest. Individual subjects’ values are reported as colored dotted lines with overlaid boxplots (lower whisker: Q1 − 1.5 * IQR; upper whisker: Q3 + 1.5 * IQR).

• 

  Figure 8. Handgrip protocol data ΔPATMd relation with ΔSBP (left), ΔDBP (center) and ΔMBP (right) in rest (green), effort (red) and recovery (blue) phases.

• 

  Figure 9. Training dataset ΔPATMd relation with ΔSBP (left), ΔDBP (center) and ΔMBP (right). Pink dots come from the handgrip protocol dataset; green dots come from the running protocol dataset.

• 

  Figure 10. Bland–Altman plots for the multiple linear regression proposed models: (a) SBP; (b) DBP; (c) MBP.

• 

  Figure 11. Estimated profiles (in blue) compared with reference device values (in red): (a) daytime; (b) nighttime.