Age-related changes in pulse risetime measured by multi-site photoplethysmography

Abstract

Objective: It is accepted that changes in the peripheral pulse waveform characteristics occur with ageing. Pulse risetime is one important feature which has clinical value. However, it is unclear how it varies across the full age spectrum from child to senior and for different peripheral measurement sites. The objectives of this study were to determine the association between age and pulse risetime characteristics over an 8-decade age range at the ears, fingers, and toes, and to consider effects arising from differences in systolic blood pressure (SBP), height and heart rate. Approach: Multi-site photoplethysmography (MPPG) pulse waveforms were recorded non-invasively from the right and left ears, fingers, and toes of 304 normal healthy human subjects (range 6–87 years; 156 male and 148 female). SBP, height, and heart rate were also measured. Multi-site PPG pulse risetimes, and their site differences, were determined. Main results: Univariate regression analysis showed positive correlations with risetime for age (ears, fingers and toes: + 0.8, + 1.9, and + 1.1 ms/year, respectively), SBP (+0.5, + 1.3, and + 0.9 ms/mmHg) and height (+0.5, + 1.2, and + 1.0 ms/cm), but with a clear inverse association with heart rate (−1.8, − 2.5, and − 1.6 ms min) (P < 0.0001). No significant differences between male and female subjects were found for pulse risetime. Significance: Normative multi-site PPG risetime characteristics have been defined in over 300 subjects and are shown to increase with age linearly up to the 8th decade. In contrast, we have shown that heart rate has a clear inverse relationship with risetime for all measurement sites.

Assessment of Non-Invasive Blood Pressure Prediction from PPG and rPPG Signals Using Deep Learning

“…Note that we did not employ any hyperparameter tuning which might further improve our results, but certainly only on a gradual scale. In contrast, other authors who did not explicitly mention a subject-based dataset split reported substantially lower prediction errors [33][34][35]62].…”

“…In a practical application, data from new subjects (e.g., from patients admitted to hospital) would lead to a prediction error much closer to the MAE reported on the non-mixed dataset rendering the whole method unsuited for clinical use. The impact of age, comorbidities, medication and measurement equipment have a large influence on PPG morphology [ 34 , 63 ]. These differences can be exploited as features for the non-invasive assessment of cardiovascular aging [ 64 , 65 ].…”

“…The variability of the pulse morphology among subjects, e.g., caused by age and cardiovascular diseases, certainly affects the association of the signal shape to a particular BP. The equipment in a hospital setting differs as well and, more importantly, the contact pressure of a PPG sensor can affect the pulse morphology [ 33 , 34 , 35 ]. Therefore the ability of the learning algorithm to generalize well may be impaired.…”

Assessment of deep learning based blood pressure prediction from PPG and rPPG signals

“…Note that we did not employ any hyperparameter tuning which might further improve our results, but certainly only on a gradual scale. In contrast, other authors who did not explicitly mention a subjectbased dataset split reported substantially lower prediction errors [30,31,32,54]. These differences suggest that morphological inter-individual variations due to age, comorbidities, medication and measurement equipment prevent the investigated NNs to generalize well.…”

“…caused by age and cardiovascular diseases, certainly affects the association of the signal shape to a particular BP. The equipment in a hospital setting differs as well and, more importantly, the contact pressure of a PPG sensor can affect the pulse morphology [30,31,32]. Therefore the ability of the learning algorithm to generalize well may be impaired.…”