Discussion of Cuffless Blood Pressure Prediction Using Plethysmograph Based on a Longitudinal Experiment: Is the Individual Model Necessary?

Figures

• 

  Figure 1. Illustration of PPG features. Inflection point of the PPG waveform between the dicrotic notch and the diastolic peak is used given that the disappearance of the diastolic peak may appear.

• 

  Figure 2. The mean absolute error (MAE) of systolic and diastolic BP of the three generalized models. The bar plot on the left corresponds to the results with scheme_1 and the bar plot on the right to the results with scheme_2. Sticks show the standard errors.

• 

  Figure 3. Comparison of three individual models using scheme_1 (left column) and scheme_2 (right column). Sticks show the standard errors.The setx means the combination of PPG features, where the set1 is {hr, a02, a25, ri}, set2 is {pir, dtr, hr, a02, a25, ri}, set3 is {hr, ri}. * means the p < 0.05 and, ** means the p < 0.001. setx_h corresponds to the scheme_1 calibration.

• 

  Figure 4. The results of the GPR_set1 (scheme_2) model. The fitting of the one-month samples of two randomly selected subjects can be seen in (a,b). The BA plot for and linear correlation can be found in (c,d).

• 

  Figure 5. Distribution of predicting errors of the PLS_set1 (left column) adn GPR_set1 models (right column). Red curves show the approximation of probability density using Gaussian distribution.

• 

  Figure 6. Distribution of similarity. Each subfigure corresponds to one subject. In each subfigure, the histogram in red shows the distribution of similarity of samples from the same subject; whereas the one in blue shows the distribution of similarity with other subjects.