The impact of healthy pregnancy on features of heart rate variability and pulse wave morphology derived from wrist-worn photoplethysmography

Figure 1. The PPG waveform and its first and second derivatives. The top waveform represents the PPG pulse wave, followed by the first and second derivatives of the waveform. Aspects of these waveforms relevant to the morphological features listed in Table 2 are indicated in the figures.

Figure 2. Ensemble plots of the PPG waveform for a pregnant and non-pregnant participant to visualize the differences between the groups. The resulting waveform is computed using the full night’s recording and the waveforms are centered (sample = 0) on the peaks.

Figure 3. Classification of pregnant and non-pregnant women based on full night recordings. (A) The AUROC scores, with standard deviation, for the models based on HRV features only (blue), morphological features only (green), and a combination of the two sets (red), plotted against the number of features used. In this case, all the segments are considered unique, disregarding the subject. (B) AUROC scores, with standard deviation, for classification based on HRV features only (blue), morphological features only (green), and a combination of the two sets (blue), stratified by participant, i.e., when the measurements of each participant are grouped, plotted against the number of features used. The features correspond to those listed in Table 3. For example, when three features are used, the first three features listed in Table 3 have been used.

Figure 4. Classification of pregnant and non-pregnant women based on data from individual sleep stages. AUROC scores, with standard deviation, for the models using only data from (A) the N1/N2 (light sleep), (B) N3 (deep sleep), or (C) REM sleep stages. For each, the results of three models are presented, based on HRV features only (blue), morphological features only (green), and a combination of the two sets (red), respectively. In (D), (E), and (F), the AUROC scores with standard deviation are again presented per sleep stage but now stratified on participant level.