Blood Pressure Estimation Using Photoplethysmography Only: Comparison between Different Machine Learning Approaches.

Pulse transit time (PTT) has been identified as a potential indicator to estimate blood pressure (BP) continuously and cufflessly in the previous studies, but it is still unsatisfactory for clinical uses. In order to improve the BP estimation accuracy, another candidate signal, photoplethysmogram (PPG) has been used to estimate continuous BP. In this study, a new PPG indicator, the intensity ratio of the first derivative wave of PPG (1st dPIR), was proposed and a novel BP estimation algorithm by using both PTT and 1st dPIR was developed with an attempt to further increase the BP estimation accuracy. The performance of the proposed algorithm was examined and validated using the ECG and PPG recordings from 22 healthy subjects when they performed both mental arithmetic stress and Valsalva's manoeuvre tests that could induce their BP fluctuations. The continuous Finapres® BP were measured with a finger cuff as a referenced BP value. Our results showed that using the proposed BP estimation algorithm, the grand absolute mean and standard deviation (SD) errors between the estimated and referenced BP were 3.22±8.02 mmHg for systolic BP (SBP) and 3.13±4.82mmHg for diastolic BP (DBP). It could achieve a decrease of 0.34±0.27 mmHg in SBP and 0.33±0.44 mmHg in DBP on the grand absolute mean and SD estimation errors in comparison to the state-of-the-art PPT-based method. Findings from this study suggest that the newly proposed PPG indicator would be promising for improving the accuracy of current PTT-based continuous and cuffless BP estimation algorithms.

Continuous blood pressure (BP) monitoring has a significant meaning to the prevention and early diagnosis of cardiovascular disease. However, existing continuous BP monitoring approaches, especially cuff-less BP monitoring approaches, are all contraptions which complex and huge computation required. For example, for the most sophisticated cuff-less BP monitoring method using pulse transit time (PTT), the simultaneous record of photoplethysmography (PPG) signal and electrocardiography (ECG) are required, and various measurement of characteristic points are needed. These issues hindered widely application of cuff less BP measurement in the wearable devices. In this study, a novel BP estimation method using single PPG signal feature was proposed and its performance in BP estimation was also tested. The results showed that the new approach proposed in this study has a mean error -0.91 ± 3.84 mmHg for SBP estimation and -0.36 ± 3.36 mmHg for DBP estimation respectively. This approach performed better than the traditional PTT based BP estimation, which mean error for SBP estimation was -0.31 ± 4.78 mmHg, and for DBP estimation was -0.18 ± 4.32 mmHg. Further investigation revealed that this new BP estimation approach only required measurement of one characteristic point, reducing much computation when implementing. These results demonstrated that this new approach might be more suitable implemented in the wearable BP monitoring devices.

A relationship between the 2017 American College of Cardiology and American Heart Association blood pressure thresholds and adverse pregnancy outcomes has been reported, but few studies have explored the diagnostic test properties of these cutoffs when used within pregnancy. Electronic databases were searched (2017-2021) for measurements of blood pressure in pregnancy at >20 weeks, classified according to the 2017 American College of Cardiology and American Heart Association criteria, and their relationship with pregnancy outcomes. Blood pressure was categorized as "normal" (systolic blood pressure of <120 mm Hg and diastolic blood pressure of <80 mm Hg), "elevated blood pressure" (systolic blood pressure of 120-129 mm Hg and diastolic blood pressure of <80 mm Hg), "stage 1 hypertension" (systolic blood pressure of 130-139 mm Hg and/or diastolic blood pressure of 80-89 mm Hg), and "stage 2 hypertension" (systolic blood pressure of ≥140 mm Hg and/or diastolic blood pressure of ≥90 mm Hg). Studies recording blood pressure at or above 20 weeks gestation were included. Meta-analyses were used to investigate the strength of the association between blood pressure cutoffs and adverse outcomes, and the diagnostic test properties were calculated accounting for gestation. There were 12 included studies. The American College of Cardiology or American Heart Association blood pressure categories were determined from peak blood pressures at any point from 20 weeks of gestation and at specific gestational ages (20-27, 28-32, or 33-36 weeks of gestation), as available. A higher (vs normal) blood pressure category was consistently associated with adverse outcomes. The strength of association between blood pressure categories and adverse outcomes was the greatest with "stage 2 hypertension" (blood pressure of ≥140/90 mm Hg). The results were similar when peak blood pressure was reported either at any time from 20 weeks of gestation or within gestational age groups (as above). No blood pressure category was useful as a diagnostic "rule-out test" for adverse outcomes, as all negative likelihood ratios were ≥0.2. Only "stage 2 hypertension" was useful as a "rule in-test," with positive likelihood ratios of ≥5.0, for maximum blood pressure at >20 weeks of gestation for preeclampsia and blood pressure within any gestational age groups for preeclampsia, eclampsia, stroke, maternal death, and stillbirth. From 20 weeks of gestation, blood pressure thresholds of 140 mm Hg (systolic) and 90 mm Hg (diastolic) were useful in identifying women at increased risk of adverse pregnancy outcomes, irrespective of the specific gestational age at blood pressure measurement. Lowering the blood pressure threshold for abnormal blood pressure at >20 weeks of gestation would not assist clinicians in identifying women at heightened maternal or perinatal risk. No American College of Cardiology or American Heart Association blood pressure threshold can provide reassurance that women are unlikely to develop adverse outcomes.

The advent of mobile devices, wearables and digital healthcare has unleashed a demand for accurate, reliable, and non-interventional ways to measure continuous blood pressure (BP). Many consumer products claim to measure BP with a cuffless device, but their lack of accuracy and reliability limit clinical adoption. Here, we demonstrate how multimodal feature datasets, comprising: (i) pulse arrival time (PAT); (ii) pulse wave morphology (PWM), and (iii) demographic data, can be combined with optimized Machine Learning (ML) algorithms to estimate Systolic BP (SBP), Diastolic BP (DBP) and Mean Arterial Pressure (MAP) within a 5 mmHg bias of the gold standard Intra-Arterial BP, well within the acceptable limits of the IEC/ANSI 80601-2-30 (2018) standard. Furthermore, DBP’s calculated using 126 datasets collected from 31 hemodynamically compromised patients had a standard deviation within 8 mmHg, while SBP’s and MAP’s exceeded these limits. Using ANOVA and Levene’s test for error means and standard deviations, we found significant differences in the various ML algorithms but found no significant differences amongst the multimodal feature datasets. Optimized ML algorithms and key multimodal features obtained from larger real-world data (RWD) sets could enable more reliable and accurate estimation of continuous BP in cuffless devices, accelerating wider clinical adoption.

The method for rapid evaluation of automatic blood pressure (BP) measurement devices (READ) is based on numerous BP measurements at rest and during a standardised postural challenge in a small number of subjects with a wide range of BPs. The present study proposes additional parameters of the READ for in-field validation of automatic BP measurement devices. BP measurements were done in supine position for 10 min followed by head-up tilt for 30 min and again supine for 10 min. BPs were determined simultaneously by automatic (AU) and mercury sphygmomanometric (MS) measurements on the same arm. The BP differences DeltaBP:AU-MS were calculated. Three units of the Colin BP-8800 and the Datex-Engstrom Cardiocaptrade mark II were evaluated. Based on DeltaBP(AU-MS), the grade of accuracy, aberration patterns and correlates of accuracy were assessed for each unit. Per unit, an average of 121 measurements were done, every BP category being met in >/=20 MS measurements. In general, the AU systolic BP values were higher than MS systolic BP values (mean systolic DeltaBP = 1.26 +/- 17.1 mm Hg) and AU diastolic BP values were lower than MS diastolic BP values (mean diastolic DeltaBP = -12.31 +/- 7.8 mm Hg). All units classified into category D of the British Hypertension Society grading system and exhibited inconsistent aberration patterns, making design of correction formulas impractical. Inaccuracy of the instruments was independent on mode of measurement, posture, magnitude of the BP and heart rate, early or late measurements from beginning of the head-up tilt test, and prolonged use of the unit. The READ permitted to identify rapidly the degree of accuracy of automatic BP measuring devices. Identification of the aberration pattern of an instrument could be the basis for calculating equations for correction of the measured BP. Further studies will show which parameters of the READ may expose specific defects of the instruments. Journal of Human Hypertension (2000) 14, 37-42.

Deep learning models for cuffless blood pressure monitoring from PPG signals using attention mechanism

Objective.Individual differences across subjects reduce the accuracy of physiological signal-based cuffless blood pressure (BP) estimation. However, training a personalized model with a large amount of data is impractical. This study aims to learn a personalized model using only a few labeled samples (e.g. 5 datapoints).Approach.This study introduced a two-stage training method to enhance the few-shot personalized model with self-supervised learning. In the first training stage, self-supervised learning is used to learn shared features from physiological signals across subjects. In the second stage, few-shot learning is used to adapt the model to each subject based on the pre-trained encoder from the first stage.Main result.Experiments were conducted on the PulseDB dataset. Under the 5-shot setting, the proposed method achieved mean absolute errors of 6.57 ± 6.22 mmHg and 3.66 ± 3.99 mmHg for systolic BP (SBP) and diastolic BP (DBP) estimation, respectively, when using photoplethysmogram (PPG) and electrocardiogram. Using only PPG signals, the method achieved 6.77 ± 6.43 mmHg and 3.80 ± 3.92 mmHg for SBP and DBP estimation, respectively. The proposed approach exceeded previous non-personalized and transfer learning methods. Its generalization capability was validated on two additional smaller datasets, demonstrating the generalization ability of the proposed method.Significant.Overall, the proposed method provides a new approach for few-shot personalization of cuffless BP estimation models, which is helpful for accurate and individualized BP estimation.

Cuffless blood pressure (BP) monitoring is a transformative technology that enables continuous noninvasive measurement of BP, which is critical for early diagnosis and prevention of cardiovascular diseases (CVD). However, current predictive models exhibit performance limitations across underrepresented demographics. This limitation is particularly pronounced in hypertensive populations, where the large BP volatility challenges accuracy of the model for BP estimation. Addressing these demographic disparities is crucial to ensure the objectivity and fairness of model for cuffless BP estimation. This paper proposes a self-organizing mixture of experts (SO-MoE) framework for BP estimation, which implements automated expert allocation mechanisms based on patient-specific characteristics. This specialized design enables precise expert selection for individualized BP estimation. We evaluated this system on the CAS-BP dataset of over seven hundred subjects and compared it with several deep learning models. The proposed system demonstrated promising performance for BP estimation with mean absolute errors (MAE) of 5.59 mmHg and 4.07 mmHg for systolic BP (SBP) and diastolic BP (DBP) respectively. Besides, the SO-MoE yielded estimation errors of -0.24 ± 7.85 mmHg for SBP and 0.33 ± 5.49 mmHg for DBP in the overall population. Key innovations of the proposed framework include reduced computational overhead, sustained temporal stability, and consistent performance across diverse demographic groups. These results establish a foundation for equitable and clinically accurate BP estimation, addressing the representation challenges inherent in cuffless BP monitoring.

Individuals with sickle cell anemia (SCA) exhibit delayed growth trajectories and lower blood pressure (BP) measurements than individuals without SCA. We evaluated factors associated with height, weight, and BP and established reference growth curves and BP tables using data from the Silent Cerebral Infarct Multi-Center Clinical (SIT) Trial (NCT00072761). Quantile regression models were used to determine the percentiles of growth and BP measurements. Multivariable quantile regression was used to test associations of baseline variables with height, weight, and BP measurements. Height and weight measurements were collected from a total of 949 participants with median age of 10.5 years [Interquartile range (IQR) 8.2-12.9] and median follow-up time of 3.2 years (IQR 1.8-4.7, range 0-12.9). Serial BP measurements were collected from a total of 944 and 943 participants, respectively, with median age of 10.6 years (IQR = 8.3-12.9 years), and median follow-up time of 3.3 years (IQR = 1.7-4.8). Multivariable quantile regression analysis revealed that higher hemoglobin measurements at baseline were associated with greater height (P < 0.001), weight (P = 0.000), systolic BP (P < 0.001), and diastolic BP (P = 0.003) measurements. We now provide new reference values for height, weight, and BP measurements that are now readily available for medical management.

Blood pressure (BP) is an important clinical vital sign that varies from beat-to-beat. Nevertheless, these variations cannot be captured by the conventional cuff-based BP monitors. This study proposes and evaluates novel cuffless frameworks to continuously estimate the 10-beat averaged systolic BP (SBP) and diastolic BP (DBP) during dynamic exercise by fusing information from multiple biosensors using five machine learning algorithms. Over 100 thousand beats of data were collected from 62 subjects (aged 59 &#x00B1; 10 years), each underwent a maximal exercise stress test. The average length of recording for each subject was 35 minutes. The BP ranges were 75-280 mmHg for SBP and 36-157 mmHg for DBP respectively. Multiple physiological parameters were measured continuously and used as inputs to five machine learning algorithms for estimating the 10-beat SBP and DBP averages before, during and after the cycling exercise. The mean absolute error (MAE) of Gaussian process regression (GPR) model was 4.8 mmHg and 3.4 mmHg for SBP and DBP, respectively. The MAE of multiple linear regression (MLR), regression tree (RT), ensemble of trees (ETs), and support vector machine (SVM) models varied from 6.1 mmHg to 17.6 mmHg and from 4.0 mmHg to 9.7 mmHg for SBP and DBP, respectively. The GPR model outperformed the other four models and showed promising results in estimating the 10-beat averages of both SBP and DBP without a cuff in a general elderly population under dynamic conditions.

Objective: Our aim was to analyze association of sodium and potassium intake on clinical course i.e. changes of blood pressure (BP) category in general population during the 7.5 years of follow up. Design and method: Out of 1582 subjects (random rural sample) enrolled in ENAH study conducted in 2008–2010, 739 were invited for the second visit in 2015 and 377 subjects with reliable both urine samples (urine creatinine in normal ranges) were included in further analyses s (80m 297w; average age 52 ± 13.5). Blood pressure was measured three times (Omron M6) and mean values were calculated. Fasting blood and first morning (spot) urine samples were collected. Sodium and potassium were determined using Tanaka, Kawasaki and Intersalt equations. Potassium intake was further adjusted for intestinal intake (77%) and renal excretion (92%). Optimal, normal, high normal BP, untreated and treated hypertension (HT) were diagnosed in 20.9%, 15.3%, 12.9%, 19.1% and 31.5%, respectively. Results: Out of 377 subjects 185 (49%) worsened (WBP) and 47 (12.5%) improved BP (IBP) category. We failed to find differences between WBP and IBP in gender, BMI and eGFR. OBP and untreated HT worsened BP and treated HT improved BP more frequently than others. At basal there were no differences in sodium and potassium intake between WBP and IBP category. At the follow up visit sodium and potassium intake were higher in WBP than in IBP group (Tanaka 175.8 ± 39.3 vs 132.8 ± 33.6; p = 0.061, 38.6 ± 8.3 vs 35.6 ± 7.4; p = 0.027) with no differences in potassium intake and Na/K ratio. Salt intake >10 g/day was significantly more frequently found in WBP than in IBP (51% vs. 34.7%; X2 = 4.918; p = 0.047). Conclusions: Worsened BP category was associated with higher salt intake. Subjects with optimal BP and untreated hypertension most frequently worsened BP and changed BP category. Very low average potassium intake (<1.5 g/day in the whole group) is plausible explanation why beneficial effect of potassium intake on BP clinical course was not observed. To lower salt intake and to increase potassium intake should be recommended to general population and not only to hypertensives.

Disentangling the phenotypic patterns of hypertension and chronic hypotension

Blood pressure (BP) categories defined by systolic BP (SBP) and diastolic BP (DBP) are commonly used. However, the BP category-specific risk of cardiovascular disease (CVD) has not been thoroughly investigated in different age groups. The aim of this study was to assess long-term CVD risk and its impact according to BP categories and age group. Pooling individual data from 10 cohorts, we studied 67 309 Japanese individuals (40-89 years old) who were free of CVD at baseline: we categorized them as belonging to three age groups: 'middle-aged' (40-64 years), 'elderly' (65-74 years) and 'very elderly' (75-89 years). BP was classified according to the 2009 Japanese Society of Hypertension Guidelines. Cox models were used to estimate adjusted hazard ratios for CVD deaths. We observed 1944 CVD deaths over a mean follow-up of 10.2 years. In all age groups, the overall relationship between BP category and CVD risk was positive, with a greater strength observed for younger age groups. We observed a trend of increased risk from SBP/DBP ≥ 130/85 mm Hg in the very elderly, and a significant increase from SBP/DBP ≥ 120/80 mm Hg in the other age groups. The population attributable fractions (PAFs) of CVD death in reference to the SBP/DBP<120/80 mm Hg category ranged from 23.4% in the very elderly to 60.3% in the middle-aged. We found an overall graded increase in CVD risk with higher BP category in the very elderly. The PAFs suggest that keeping BP levels low is an important strategy for primary CVD prevention, even in an elderly population.

BackgroundBlood pressure is a major modifiable risk factor for both cardiovascular disease and dementia, yet its relationship with blood-based biomarkers of Alzheimer's disease (AD) and neurodegeneration remains unclear.ObjectiveTo evaluate associations of systolic blood pressure, diastolic blood pressure (DBP), and pulse pressure with plasma biomarkers of AD and neurodegeneration, and to examine whether these associations differ across Non-Hispanic White, Non-Hispanic Black, and Hispanic participants.MethodsParticipants were 2636 cognitively normal older adults enrolled in the Health and Aging Brain Study-Health Disparities. Participants with hypotension or hypertensive crisis were excluded. The outcomes were plasma biomarkers including phosphorylated Tau181, phosphorylated tau217 (p-Tau217), amyloid-β 42 to 40 ratio, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP). Adjusted multivariable linear regression models evaluated associations between blood pressure measures and log transformed plasma biomarker concentrations. Additional models examined the American Heart Association blood pressure categories as a predictor.ResultsHigher DBP was associated with lower concentrations of neurodegeneration related plasma biomarkers, including p-Tau217, NfL, and GFAP, after correction for multiple comparisons. Analyses using American Heart Association blood pressure categories demonstrated that stage 1 and stage 2 hypertension were associated with lower GFAP concentrations compared with normal blood pressure, with patterns that were broadly consistent across racial and ethnic subgroups.ConclusionsIn this multi-racial ethnic cohort of cognitively normal older adults without hypotension or hypertensive crisis, higher DBP and clinical blood pressure categories were associated with lower levels of select plasma biomarkers of neurodegeneration (NfL and GFAP).

Mercury-free sphygmomanometers are gradually replacing the traditional sphygmomanometers in clinical settings and epidemiological surveys for measuring blood pressure (BP) due to mercury toxicity. No direct comparative studies have evaluated BP differences and statistical errors of automated oscillometric devices (ODs) against electronic auscultatory devices (ADs) for epidemiologic surveys. Herein, we evaluated the validity of ODs for the Korea National Health and Nutrition Examination Survey (KNHANES) using the Universal Standard for BP device validation through a direct comparison with ADs as the reference standard. Four trained observers performed validation on 278 volunteers aged ≥ 19 years with a standardized BP measurement protocol. Agreement between the BP measurements recorded with an OD against those recorded with an AD was assessed by Lin’s concordance correlation coefficient (CCC) and Bland–Altman’s limits of agreement. To evaluate the agreement for BP classification, weighted kappa values were estimated. To explore the factors associated with BP measurement differences between the 2 devices, multiple linear regression analysis was performed. The average BP differences (OD-AD) were 2.6 ± 6.2 mm Hg for systolic BP (SBP) and −5.1 ± 5.6 mm Hg for diastolic BP (DBP). Lin’s CCCs were 0.927 and 0.768 for the overall SBP and DBP, respectively. The cumulative percentage of absolute errors ≤10 mm Hg was 88.1% for SBP and 81.3% for DBP. The weighted kappa value for the Joint National Committee 7 BP classification was 0.75 (95% confidence interval: 0.68–0.81). An OD overestimated the prevalence of SBP (0.3%, P = .0222) and underestimated the prevalence of DBP (1.8%, P < .0001). Multivariate analysis to identify the risk factors for BP difference revealed the arm circumference (AC) to be negatively associated with BP difference. Male sex was positively associated, while age was negatively associated with SBP difference. OD-DBP was positively associated with DBP difference and negatively associated for DBP absolute error. ODs met the accuracy requirements of the Universal Standard criteria against ADs for SBP but not for DBP. Thus, the DBP values may be underestimated by ODs in the KNHANES.