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

Arterial stiffness assessment using PPG feature extraction and significance testing in an in vitro cardiovascular system

Cardiovascular diseases (CVDs) remain the leading cause of global mortality, therefore understanding arterial stiffness is essential to developing innovative technologies to detect, monitor and treat them. The ubiquitous spread of photoplethysmography (PPG), a completely non-invasive blood-volume sensing technology suitable for all ages, highlights immense potential for arterial stiffness assessment in the wider healthcare setting outside specialist clinics, for example during routine visits to a General Practitioner or even at home with the use of mobile and wearable health devices. This study employs a custom-manufactured in vitro cardiovascular system with vessels of varying stiffness to test the hypothesis that PPG signals may be used to detect and assess the level of arterial stiffness under controlled conditions. Analysis of various morphological features demonstrated significant (p < 0.05) correlations with vessel stiffness. Particularly, area related features were closely linked to stiffness in red PPG signals, while for infrared PPG signals the most correlated features were related to pulse-width. This study demonstrates the utility of custom vessels and in vitro investigations to work towards non-invasive cardiovascular assessment using PPG, a valuable tool with applications in clinical healthcare, wearable health devices and beyond.

Insights into vascular physiology from sleep photoplethysmography

STUDY OBJECTIVES

Photoplethysmography (PPG) in consumer sleep trackers is now widely available and used to assess heart rate variability (HRV) for sleep staging. However, PPG waveform changes during sleep can also inform about vascular elasticity in healthy persons who constitute a majority of users. To assess its potential value, we traced the evolution of PPG pulse waveform during sleep alongside measurements of HRV and blood pressure (BP).

METHODS

Seventy-eight healthy adults (50% male, median [IQR range] age: 29.5 [23.0, 43.8]) underwent overnight polysomnography (PSG) with fingertip PPG, ambulatory blood pressure monitoring, and electrocardiography (ECG). Selected PPG features that reflect arterial stiffness: systolic to diastolic distance (∆T_norm), normalized rising slope (Rslope) and normalized reflection index (RI) were derived using a custom-built algorithm. Pulse arrival time (PAT) was calculated using ECG and PPG signals. The effect of sleep stage on these measures of arterial elasticity and how this pattern of sleep stage evolution differed with participant age were investigated.

RESULTS

BP, heart rate (HR) and PAT were reduced with deeper non-REM sleep but these changes were unaffected by the age range tested. After adjusting for lowered HR, ∆T_norm, Rslope, and RI showed significant effects of sleep stage, whereby deeper sleep was associated with lower arterial stiffness. Age was significantly correlated with the amount of sleep-related change in ∆T_norm, Rslope, and RI, and remained a significant predictor of RI after adjustment for sex, body mass index, office BP, and sleep efficiency.

CONCLUSIONS

The current findings indicate that the magnitude of sleep-related change in PPG waveform can provide useful information about vascular elasticity and age effects on this in healthy adults.

Nocturnal Blood Pressure Estimation from Sleep Plethysmography Using Machine Learning

BACKGROUND

Elevated nocturnal blood pressure (BP) is a risk factor for cardiovascular disease (CVD) and mortality. Cuffless BP assessment aided by machine learning could be a desirable alternative to traditional cuff-based methods for monitoring BP during sleep. We describe a machine-learning-based algorithm for predicting nocturnal BP using single-channel fingertip plethysmography (PPG) in healthy adults.

METHODS

Sixty-eight healthy adults with no apparent sleep or CVD (53% male), with a median (IQR) age of 29 (23-46 years), underwent overnight polysomnography (PSG) with fingertip PPG and ambulatory blood pressure monitoring (ABPM). Features based on pulse morphology were extracted from the PPG waveforms. Random forest models were used to predict night-time systolic blood pressure (SBP) and diastolic blood pressure (DBP).

RESULTS

Our model achieved the highest out-of-sample performance with a window length of 7 s across window lengths explored (60 s, 30 s, 15 s, 7 s, and 3 s). The mean absolute error (MAE ± STD) was 5.72 ± 4.51 mmHg for SBP and 4.52 ± 3.60 mmHg for DBP. Similarly, the root mean square error (RMSE ± STD) was 6.47 ± 1.88 mmHg for SBP and 4.62 ± 1.17 mmHg for DBP. The mean correlation coefficient between measured and predicted values was 0.87 for SBP and 0.86 for DBP. Based on Shapley additive explanation (SHAP) values, the most important PPG waveform feature was the stiffness index, a marker that reflects the change in arterial stiffness.

CONCLUSION

Our results highlight the potential of machine learning-based nocturnal BP prediction using single-channel fingertip PPG in healthy adults. The accuracy of the predictions demonstrated that our cuffless method was able to capture the dynamic and complex relationship between PPG waveform characteristics and BP during sleep, which may provide a scalable, convenient, economical, and non-invasive means to continuously monitor blood pressure.

Deep learning classification of systemic sclerosis from multi-site photoplethysmography signals

Introduction: A pilot study assessing a novel approach to identify patients with Systemic Sclerosis (SSc) using deep learning analysis of multi-site photoplethysmography (PPG) waveforms ("DL-PPG"). Methods: PPG recordings having baseline, unilateral arm pressure cuff occlusion and reactive hyperaemia flush phases from 6 body sites were studied in 51 Controls and 20 SSc patients. RGB scalogram images were obtained from the PPG, using the continuous wavelet transform (CWT). 2 different pre-trained convolutional neural networks (CNNs, namely, GoogLeNet and EfficientNetB0) were trained to classify the SSc and Control groups, evaluating their performance using 10-fold stratified cross validation (CV). Their classification performance (i.e., accuracy, sensitivity, and specificity, with 95% confidence intervals) was also compared to traditional machine learning (ML), i.e., Linear Discriminant Analysis (LDA) and K-Nearest Neighbour (KNN). Results: On a participant basis DL-PPG accuracy, sensitivity and specificity for GoogLeNet were 83.1 (72.3-90.9), 75.0 (50.9-91.3) and 86.3 (73.7-94.3)% respectively, and for EfficientNetB0 were 87.3 (77.2-94.0), 80.0 (56.3-94.3) and 90.1 (78.6-96.7)%. The corresponding results for ML classification using LDA were 66.2 (53.9-77.0), 65.0 (40.8-84.6) and 66.7 (52.1-79.2)% respectively, and for KNN were 76.1 (64.5-85.4), 40.0 (19.1-63.9), and 90.2 (78.6-96.7)% respectively. Discussion: This study shows the potential of DL-PPG classification using CNNs to detect SSc. EfficientNetB0 gave an overall improved performance compared to GoogLeNet, with both CNNs performing better than the traditional ML methods tested. Our automatic AI approach, using transfer learning, could offer significant benefits for SSc diagnostics in a variety of clinical settings where low-cost portable and easy-to-use diagnostics can be beneficial.

Filtering-induced changes of pulse transmit time across different ages: a neglected concern in photoplethysmography-based cuffless blood pressure measurement

Background: Pulse transit time (PTT) is a key parameter in cuffless blood pressure measurement based on photoplethysmography (PPG) signals. In wearable PPG sensors, raw PPG signals are filtered, which can change the timing of PPG waveform feature points, leading to inaccurate PTT estimation. There is a lack of comprehensive investigation of filtering-induced PTT changes in subjects with different ages. Objective: This study aimed to quantitatively investigate the effects of aging and PTT definition on the infinite impulse response (IIR) filtering-induced PTT changes. Methods: One hundred healthy subjects in five different ranges of age (i.e., 20-29, 30-39, 40-49, 50-59, and over 60 years old, 20 subjects in each) were recruited. Electrocardiogram (ECG) and PPG signals were recorded simultaneously for 120 s. PTT was calculated from the R wave of ECG and PPG waveform features. Eight PTT definitions were developed from different PPG waveform feature points. The raw PPG signals were preprocessed then further low-pass filtered. The difference between PTTs derived from preprocessed and filtered PPG signals, and the relative difference, were calculated and compared among five age groups and eight PTT definitions using the analysis of variance (ANOVA) or Scheirer-Ray-Hare test with post hoc analysis. Linear regression analysis was used to investigate the relationship between age and filtering-induced PTT changes. Results: Filtering-induced PTT difference and the relative difference were significantly influenced by age and PTT definition (p < 0.001 for both). Aging effect on filtering-induced PTT changes was consecutive with a monotonous trend under all PTT definitions. The age groups with maximum and minimum filtering-induced PTT changes depended on the definition. In all subjects, the PTT defined by maximum peak of PPG had the minimum filtering-induced PTT changes (mean: 16.16 ms and 5.65% for PTT difference and relative difference). The changes of PTT defined by maximum first PPG derivative had the strongest linear relationship with age (R-squared: 0.47 and 0.46 for PTT difference relative difference). Conclusion: The filtering-induced PTT changes are significantly influenced by age and PTT definition. These factors deserve further consideration to improve the accuracy of PPG-based cuffless blood pressure measurement using wearable sensors.

Using the photoplethysmography method to monitor age-related changes in the cardiovascular system

Introduction: Aging is a physiological process characterized by progressive changes in all organ systems. In the last few decades, the elderly population has been growing, so the scientific community is focusing on the investigation of the aging process, all in order to improve the quality of life in elderly. One of the biggest challenges in studying the impact of the aging on the human body represents the monitoring of the changes that inevitably occur in arterial blood vessels. Therefore, the medical community has invested a great deal of effort in studying and discovering new methods and tools that could be used to monitor the changes in arterial blood vessels caused by the aging process. The goal of our research was to develop a new diagnostic method using a photoplethysmographic sensor and to examine the impact of the aging process on the cardiovascular system in adults. Long-term recorded arterial blood flow waveforms were analyzed using detrended fluctuation analysis. Materials and Methods: The study included 117 respondents, aged 20-70 years. The waveform of the arterial blood flow was recorded for 5 min, with an optical sensor placed above the left common carotid artery, simultaneously with a single-channel ECG. For each cardiac cycle, the blood flow amplitude was determined, and a new time series was formed, which was analyzed non-linearly (DFA method). The values of the scalar coefficients α 1 and α 2, particularly their ratio (α 1/α 2) were obtained, which were then monitored in relation to the age of the subjects. Result: The values of the scalar ratio (α 1/α 2) were significantly different between the subjects older and younger than 50 years. The value of the α 1/α 2 decreased exponentially with the aging. In the population of middle-aged adults, this ratio had a value around 1, in young adults the value was exclusively higher than 1 and in older adults the value was exclusively lower than 1. Conclusion: The results of this study indicated that the aging led to a decrease in the α 1/α 2 in the population of healthy subjects. With this non-invasive method, changes in the cardiovascular system due to aging can be detected and monitored.

A Computational Modeling and Simulation Workflow to Investigate the Impact of Patient-Specific and Device Factors on Hemodynamic Measurements from Non-Invasive Photoplethysmography

Cardiovascular disease is the leading cause of death globally. To provide continuous monitoring of blood pressure (BP), a parameter which has shown to improve health outcomes when monitored closely, many groups are trying to measure blood pressure via noninvasive photoplethysmography (PPG). However, the PPG waveform is subject to variation as a function of patient-specific and device factors and thus a platform to enable the evaluation of these factors on the PPG waveform and subsequent hemodynamic parameter prediction would enable device development. Here, we present a computational workflow that combines Monte Carlo modeling (MC), gaussian combination, and additive noise to create synthetic dataset of volar fingertip PPG waveforms representative of a diverse cohort. First, MC is used to determine PPG amplitude across age, skin tone, and device wavelength. Then, gaussian combination generates accurate PPG waveforms, and signal processing enables data filtration and feature extraction. We improve the limitations of current synthetic PPG frameworks by enabling inclusion of physiological and anatomical effects from body site, skin tone, and age. We then show how the datasets can be used to examine effects of device characteristics such as wavelength, analog to digital converter specifications, filtering method, and feature extraction. Lastly, we demonstrate the use of this framework to show the insensitivity of a support vector machine predictive algorithm compared to a neural network and bagged trees algorithm.

XGBoost Regression of the Most Significant Photoplethysmogram Features for Assessing Vascular Aging

The purpose of this study was to confirm the potential of XGBoost as a vascular aging assessment model based on the photoplethysmogram (PPG) features suggested in previous studies, and to explore the key PPG features for vascular aging assessment through an explainable artificial intelligence method. The PPG waveforms obtained from 752 volunteers aged 19-87 years were analyzed and a total of 78 features were derived that were proposed in previous studies. Age was estimated through an XGBoost regression model, and estimation error was calculated in terms of mean absolute error and root-mean-squared error. To evaluate feature importance, gain, coverage, weight, and SHAP value was calculated. The vascular aging assessment model developed using XGBoost has 8.1 years of mean-absolute error and 9.9 years of root-mean-squared error, a correlation coefficient of 0.63 with actual age, and a coefficient of determination of 0.39. Feature importance analysis using the SHAP value confirmed that features, such as systolic and diastolic peak amplitude, risetime, skewness, and pulse area, play a key role in vascular aging assessment. The XGBoost regression model showed an equal level of performance to the existing PPG-based vascular aging assessment models. Moreover, the result of feature importance analysis using explainable artificial intelligence verified that the features proposed in previous vascular aging assessment studies, such as reflective index and risetime, were more important in vascular aging assessment than other PPG features.

Assessing hemodynamics from the photoplethysmogram to gain insights into vascular age: A review from VascAgeNet

The photoplethysmogram (PPG) signal is widely measured by clinical and consumer devices, and it is emerging as a potential tool for assessing vascular age. The shape and timing of the PPG pulse wave are both influenced by normal vascular aging, changes in arterial stiffness and blood pressure, and atherosclerosis. This review summarizes research into assessing vascular age from the PPG. Three categories of approaches are described: 1) those which use a single PPG signal (based on pulse wave analysis), 2) those which use multiple PPG signals (such as pulse transit time measurement), and 3) those which use PPG and other signals (such as pulse arrival time measurement). Evidence is then presented on the performance, repeatability and reproducibility, and clinical utility of PPG-derived parameters of vascular age. Finally, the review outlines key directions for future research to realize the full potential of photoplethysmography for assessing vascular age.

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

Exploiting photoplethysmography signals (PPG) for non-invasive blood pressure (BP) measurement is interesting for various reasons. First, PPG can easily be measured using fingerclip sensors. Second, camera based approaches allow to derive remote PPG (rPPG) signals similar to PPG and therefore provide the opportunity for non-invasive measurements of BP. Various methods relying on machine learning techniques have recently been published. Performances are often reported as the mean average error (MAE) on the data which is problematic. This work aims to analyze the PPG- and rPPG based BP prediction error with respect to the underlying data distribution. First, we train established neural network (NN) architectures and derive an appropriate parameterization of input segments drawn from continuous PPG signals. Second, we use this parameterization to train NNs with a larger PPG dataset and carry out a systematic evaluation of the predicted blood pressure. The analysis revealed a strong systematic increase of the prediction error towards less frequent BP values across NN architectures. Moreover, we tested different train/test set split configurations which underpin the importance of a careful subject-aware dataset assignment to prevent overly optimistic results. Third, we use transfer learning to train the NNs for rPPG based BP prediction. The resulting performances are similar to the PPG-only case. Finally, we apply different personalization techniques and retrain our NNs with subject-specific data for both the PPG-only and rPPG case. Whilst the particular technique is less important, personalization reduces the prediction errors significantly.