Arterial Blood Pressure Signals Research Articles

Reconstruction of arterial blood pressure waveforms based on squeeze-and-excitation network models using electrocardiography and photoplethysmography signals

Arterial blood pressure (ABP) waveforms indicate the efficiency with which a patient’s blood responds to changes in the arterial flow. ABP waveforms can be used to analyze the cardiovascular health of patients, including heart rate analysis, stress analysis, and cardiovascular disease diagnosis. However, obtaining the ABP waveform is more difficult than recording the electrocardiography (ECG) and photoplethysmography (PPG) signals. Several studies have focused on the reconstruction of the ABP waveform by recombining bio-signals. Reconstructing ABP waveforms is considered a major challenge because of signals being distorted by waveform destruction, owing to the normalization and standardization of the input ABP waveform. This paper presents a model that can simultaneously output blood pressure estimation and ABP waveforms using the raw ABP signal as the input. Herein, the squeeze-and-excitation network (SE-Net) is coupled with the multi-task learning architecture. Three signals, namely, the ECG, PPG, and ECG–PPG, were trained as feature vectors using different variants of the U-Net model, and the results are compared. Additionally, we introduce a novel method for incorporating white Gaussian noise at different signal-to-noise ratios (SNRs) to augment the training data. The objective of the proposed approach is to evaluate the ABP signal reconstruction capability of the SE-Net at various SNR levels and analyze the overall performance and robustness of the model. In the case of systolic blood pressure, the root mean square error (RMSE) of the PPG-based values is 2.58 mmHg, and the Pearson correlation coefficient is r = 0.92 (p≤ 0.001). In the case of diastolic blood pressure(DBP), the RMSE was 3.35 mmHg, and the Pearson correlation coefficient was r = 0.89 (p ≤ 0.001). The results indicate the potential for replacing ECG signals with PPG signals to overcome the constraints of optimizing the non-invasive reconstruction of ABP waveforms using SE-Net. Furthermore, we demonstrate that the SE-Net-combined multi-task learning architecture model can simultaneously perform blood pressure estimation and output ABP waveforms without damaging the raw signal.

Estimation of Systolic and Diastolic Blood Pressure for Hypertension Identification from Photoplethysmography Signals

Continuous monitoring plays a crucial role in diagnosing hypertension, characterized by the increase in Arterial Blood Pressure (ABP). The gold-standard method for obtaining ABP involves the uncomfortable and invasive technique of cannulation. Conversely, ABP can be acquired non-invasively by using Photoplethysmography (PPG). This non-invasive approach offers the advantage of continuous BP monitoring outside a hospital setting and can be implemented in cost-effective wearable devices. PPG and ABP signals differ in scale values, which creates a non-linear relationship, opening avenues for the utilization of algorithms capable of detecting non-linear associations. In this study, we introduce Neural Model of Blood Pressure (NeuBP), which estimates systolic and diastolic values from PPG signals. The problem is treated as a binary classification task, distinguishing between Normotensive and Hypertensive categories. Furthermore, our research investigates NeuBP’s performance in classifying different BP categories, including Normotensive, Prehypertensive, Grade 1 Hypertensive, and Grade 2 Hypertensive cases. We evaluate our proposed method by using data from the publicly available MIMIC-III database. The experimental results demonstrate that NeuBP achieves results comparable to more complex models with fewer parameters. The mean absolute errors for systolic and diastolic values are 5.02 mmHg and 3.11 mmHg, respectively.

An end-to-end deep learning framework for accurate estimation of intracranial pressure waveform characteristics

It is clinically significant if intracranial pressure (ICP) waveform could be reliably reconstructed from simple physiological measurements that are commonly used in clinical practice. Some attempts have been made with the focus on estimating intermittent mean ICP (MICP) values, but no investigation has been performed to reconstruct beat-by-beat ICP waveforms from arterial blood pressure (ABP) waveforms. This study aimed to overcome this challenge. A large dataset of 28511 segments (10.8 s for each segment; 540 samples) of simultaneously measured invasive ABP and ICP signals was used to develop and cross-evaluate a Wave-U-Net based deep learning model for reconstructing ICP waveform from ABP signals. From the reconstructed beat-by-beat ICP waveform, systolic ICP, diastolic ICP, and MICP were derived for each cardiac cycle and compared with reference ICPs. After comparing all the sampling points of each segment, the results showed that the reconstructed ICP waveforms were statistically correlated to the reference ICP waveforms, with the average Pearson's correlation coefficient (r) of 0.822 across all the segments. There was no significant difference between the estimated and reference ICP values, with the mean absolute error (MAE) of systolic ICP, diastolic ICP and MICP of 0.75 ± 0.29, 0.48 ± 0.22 and 0.42 ± 0.18 mmHg, respectively, meeting the clinical standards of ICP monitoring for the management of traumatic brain injury. Our proposed method confirmed the hypothesis for the first time that ABP waveform can be used for reliable reconstruction of ICP waveform, enabling continuous monitoring of ICP waveform, and providing the hope possibility to replace the gold standard invasive technique.

Malignant arrhythmias detection using a synthesis-by-analysis modeling method of arterial blood pressure signal

Extreme bradycardia, extreme tachycardia, ventricular flutter fib, and ventricular tachycardia are four malignant arrhythmias (MAs) that lead to sudden cardiac death. It is very important to detect them in daily life. The arterial blood pressure (ABP) signal contains abundant pathological information about four MAs and is easy to be recorded under domestic conditions. Thus, a synthesis-by-analysis (SA) modeling method for ABP signal was proposed to detect the four MAs in this study. The average models of MAs and healthy subjects were obtained by SA modeling, and the change of each ABP wave was quantitively described by twelve parameters of wave models. Then, the probabilistic neural network (PNN) and random forest (RF) are trained to detect the MAs. The experimental data were employed from Fantasia and the 2015 PhysioNet/CinC Challenge databases. The SA modeling results show that some pathological and physiological changes could be extracted from the average models. The two-sample ks-test results between different groups are markedly different (h = 1, p < 0.05). The detection results show that the performances of PPN classifiers are less than that of RF. The kappa coefficients (KC) for the RF classifiers are 97.167 ± 1.46 %, 97.888 ± 0.808 %, 99.895 ± 0.545 %, 98.575 ± 1.683 % and 92.241 ± 1.517 %, respectively. The mean KC is 97.083 ± 0.67 %. Compared to the performance of some existing studies, the proposed method has better performance and is potential to diagnose MAs in m-health.

Peripheral artery disease diagnosis based on deep learning-enabled analysis of non-invasive arterial pulse waveforms

This paper intends to investigate the feasibility of peripheral artery disease (PAD) diagnosis based on the analysis of non-invasive arterial pulse waveforms. We generated realistic synthetic arterial blood pressure (BP) and pulse volume recording (PVR) waveform signals pertaining to PAD present at the abdominal aorta with a wide range of severity levels using a mathematical model that simulates arterial blood circulation and arterial BP-PVR relationships. We developed a deep learning (DL)-enabled algorithm that can diagnose PAD by analyzing brachial and tibial PVR waveforms, and evaluated its efficacy in comparison with the same DL-enabled algorithm based on brachial and tibial arterial BP waveforms as well as the ankle-brachial index (ABI). The results suggested that it is possible to detect PAD based on DL-enabled PVR waveform analysis with adequate accuracy, and its detection efficacy is close to when arterial BP is used (positive and negative predictive values at 40 % abdominal aorta occlusion: 0.78 vs 0.89 and 0.85 vs 0.94; area under the ROC curve (AUC): 0.90 vs 0.97). On the other hand, its efficacy in estimating PAD severity level is not as good as when arterial BP is used (r value: 0.77 vs 0.93; Bland-Altman limits of agreement: −32%-+32 % vs −20%-+19 %). In addition, DL-enabled PVR waveform analysis significantly outperformed ABI in both detection and severity estimation. In sum, the findings from this paper suggest the potential of DL-enabled non-invasive arterial pulse waveform analysis as an affordable and non-invasive means for PAD diagnosis.

Worse glycemic control is associated with an impaired arterial energy transfer pattern in elderly with diabetes

Abstract Background Worse glycaemic control is known to exacerbate endothelial insult, increasing macro- and microvascular complication risks in diabetes. Impaired intracranial haemodynamics have been frequently reported in patients with diabetes alongside with malperfusion causing structural abnormalities, and poor cognitive and cortical functionality (1-2).Assessment of the alterations in energy transfer characteristics derived from hemodynamic signals may provide novel mechanistic information about how worse glycaemic control may affect cerebrovascular perfusion. Wave reflection index(WRI) is a previously proposed measure of arterial energy transfer efficiency which reflects distal impedance mismatch(3-4). Purpose We aim to interrogate potential relationship between a worse glycaemic control and a less efficient arterial energy transfer, which could enhance our understanding of the cerebrovascular-cortical functional coupling on a mechanistic basis in elderly with diabetes. Methods Left middle cerebral artery wave intensity (WI) was computed with arterial blood pressure and flow velocity (transcranial Doppler) signal as previously described (WI = (dP/dt)*(dU/dt)) (Figure 1). Standard statistical tests were used to interrogate the relationship between glycated haemoglobin levels (HbA1c) and WI parameters. Results In 36 patients with diabetes mellitus type-2 (DM2) (mean age = 65.7 ± 8.8 years), backward compression wave intensity (BCWI) and wave reflection ratio (WRI= BCWI/FCWI) were significantly correlated with HbA1c levels before (rBCWI: 0.409 pBCWI:0.013; rWRI:0.345 pWRI:0.040) and after (rBCWI: 0.410 pBCWI:0.018, rWRI:0.351 pWRI:0.045) adjusting for age, sex and mean arterial pressure (Figure 2). Conclusion(s) In elderly patient with DM2, higher HbA1c values indicating worse glycaemic control are significantly correlated with altered cerebrovascular energy transfer characteristics as assessed by MCA wave intensity analysis. Continuous relationship between worse glycaemic control and higher wave reflection index (a measure of less efficient energy transfer in arterial system and distal impedance mismatch) may imply that the use of energy transfer related indices alongside with conventional hemodynamic indices may help providing a more comprehensive pathophysiological understanding about how cardiovascular risk factors interfere with perfusion.Computation of Wave IntensityMain Results

Abstract 189: How Quickly Does Arterial Blood Pressure Recover After Pauses in Mechanical Chest Compressions in Humans With Out-of-Hospital Cardiac Arrest?

Introduction: Conventional wisdom says that arterial blood pressure (ABP) recovers slowly after interrupting CPR chest compressions (CCs). An oft-cited animal study with manual compressions and 15:2 compression:ventilation ratio showed that all 15 compressions were needed to restore pre-pause invasive pressure, implying that even short pauses substantially reduce CPR blood flow. Whether this holds true for humans receiving mechanical chest compressions is unknown. We measured recovery of ABP after pauses in mechanical compressions by LUCAS in patients treated for out-of-hospital cardiac arrest (OHCA). Methods: We analyzed data from a subset of patients enrolled in both groups of a prospective, randomized LUCAS 2 Active Decompression trial performed by an anesthesiologist-staffed rapid response car program in Oslo, Norway from 2016 to 2018. Only patients with an ABP signal and at least one CC pause longer than 2 sec were included. All CCs were mechanical. Arterial cannulation was verified, and CC pauses were identified by physician review of the patient care record and physiological signals. Pauses were excluded if ROSC or a LUCAS “soft start” occurred or epinephrine had been applied within 5 minutes of the pause. Peak compression (cAP), mean (mAP), and mean decompression arterial blood pressures (dAP) for 10 CCs before and after each pause were measured with custom MATLAB code. Results: We included 56 patients with a total of 148 pauses (mean pause duration: 10.63 ± 3.05 sec). Pre-pause mean cAP, mAP, and dAP were 95.6 ± 0.3, 49.9 ± 0.1 and 30.5 ± 0.1 mmHg. Post-pause dAP for each compression was not significantly different than pre-pause dAP (Fig. 1A,B). cAP and mAP recovered to &gt;90% of pre-pause pressure within 2 CCs. Pause duration was not predictive of post-pause pressures (Fig. 1C). Conclusions: ABP generated by mechanical CPR recovered quickly after pauses. Consequently, short pauses (such as pauses for 30:2 CPR) may pose less hemodynamic concern than previously thought.

Statistical shape modeling-based algorithm for replacing missing beats in blood pressure signals.

In clinical environments, such as the Intensive Care Unit (ICU), continuous and uninterrupted monitoring of vital signs is critical for the early detection of patient deterioration and prompt intervention. Since data collected in these settings are often corrupted by noise, artifacts, or recording gaps, it is important to estimate missing data for a more accurate signal assessment.In this study, we propose an automatic algorithm for reconstructing of arterial blood pressure signal waveforms. The methodological core of the algorithm is based on the idea of statistical shape modeling, which basically estimates the shape variation of beat waveforms in order to reconstruct them in noisy segments. The waveform reconstruction is achieved by combining the average beat template from a 90-second segment of clean signal preceding the gap with the main shape variations of the estimated waveform.The algorithm was validated using arterial blood pressure recordings from 9 subjects admitted in the ICU and collected in the MIMIC-III Waveform Database, each lasting 1 hour and sampled at 125 Hz. For each record, ten fictitious gaps were created, and the reconstructed segments were compared to the original signals with the metrics proposed within the PhysioNet / Computing in Cardiology Challenge 2010. Results demonstrate the excellent performance of the proposed algorithm, with overall averages of both Q1 and Q2 metrics greater than 0.85.

A novel CS-NET architecture based on the unification of CNN, SVM and super-resolution spectrogram to monitor and classify blood pressure using photoplethysmography

ContextContinuous blood pressure (BP) monitoring plays an important role while treating various cardiovascular diseases and hypertension. A high correlation between arterial blood pressure (ABP) and Photoplethysmogram (PPG) signal enables using a PPG signal to monitor and classify BP continuously. Control of BP in realtime is the basis for the prevention of hypertension. Proposed approachThis work proposes a CS-NET architecture by unifying CNN and SVM approaches to classify BP using PPG signals. The main objective of the CS-NET method is to establish an accurate and reliable algorithm for the ABP classification. MethodologyABP signals are labeled normal and abnormal using the hypertension criteria the American College of Cardiology (ACC)/American Heart Association (AHA) laid down. The proposed CS-NET model incorporates three critical steps in three successive stages. The first stage includes converting a preprocessed PPG signal into a time-frequency (TF) representation called a super-resolution spectrogram by superlet transform. The second stage uses a convolutional neural network (CNN) model with several hidden layers to extract morphological features from every PPG super-resolution spectrogram. The third stage uses a support vector machine (SVM) classifier to classify the PPG signal. ResultsPPG signals are used to train and test the proposed model. The performance of the proposed CS-NET method is tested using MIMIC-II, MIMIC-III, and PPG-BP-figshare database in terms of accuracy and F1 score. Moreover, the CS-NET method achieves better results with high accuracy when compared with other benchmark approaches that require an electrocardiogram signal for reference. ConclusionsThe proposed model achieved an aggregate classification accuracy of 98.21% across a five-fold cross-validation technique, making it a reliable approach for BP classification in clinical settings and realtime monitoring.

Modeling peripheral arterial and venous pressure signals with integral pulse frequency modulation

This paper proposes a novel mathematical modeling framework for peripheral arterial and peripheral venous blood pressure signals from porcine experiments. Peripheral blood pressure signals can be acquired using regular catheters during standard patient treatment. The minimally invasive nature and ubiquitous availability of catheters render it an ideal candidate for various applications. However, there is no analytical model for peripheral blood pressure signals in the literature. We address this issue by proposing a model for these signals under the integral pulse frequency modulation (IPFM) framework. The model incorporates the impacts of physiological phenomena, such as the heartbeat pulse shape variation, heart rate variability, respiratory rate, etc. The model parameters are obtained by applying the IPFM model to experimental data collected from four pigs under different anesthetic dosages. The proposed model can fit the experimental data with Pearson correlation coefficients greater than 0.99 and 0.90 for arterial and venous blood pressure signals, respectively. The performance of model-synthesized data on the classification of two different anesthesia is comparable with experimental data. Parameters like pulse shape and duration can also work as distinguishable features under different anesthesia. We also proposed a way to distinguish respiratory-induced heart rate variability from other causes. Increasing doses of vasodilating anesthesia is similar to going from dehydration to hydration. Thus the results obtained in this study can be extended in distinguishing hydrated and dehydrated subjects. This model can be extended to similar biomedical signals like photoplethysmography, cerebral blood flow velocity, and Doppler waveforms.

Hybrid modeling on reconstitution of continuous arterial blood pressure using finger photoplethysmography

The continuous estimation of arterial blood pressure (ABP) waveforms directly from single-channel photoplethysmography (PPG) signals will predictably change the way to monitor blood pressure in the future. This article proposed a new hybrid mathematical model for continuous blood pressure monitoring by investigating the relationship between the finger PPG signal and the radial ABP signal based on a public database. Considering potential damping factors and wave propagation/reflection in blood circulation, we combined the electrical network model with the tube-load model. The optimal range of model parameters was obtained through the system identification method to realize the individualized continuous blood pressure measurements. Compared with the invasive measurement, the hybrid model performed superior blood pressure estimation with high consistency. The estimated ABP waveforms correlated highly with the reference waveforms with an average correlation coefficient of 0.96. The mean absolute error/standard deviation of the estimated systolic blood pressure (SBP), mean arterial blood pressure (MAP), and diastolic blood pressure (DBP) were 3.0/4.4, 2.1/3.0 and 2.1/3.2 mmHg, respectively. The results met the requirements of the Association for the Advancement of Medical Instrumentation (AAMI). The hybrid model is expected to be embedded in small wearable devices to directly estimate the continuous blood pressure waveforms at the radial artery site through the PPG signals, pioneering the synchronous non-sensing monitoring on blood pressure and blood flow.

Beat-to-beat finger photoplethysmography in atrial fibrillation patients undergoing electrical cardioversion

Atrial fibrillation (AF)-induced peripheral microcirculatory alterations have poorly been investigated. The present study aims to expand current knowledge through a beat-to-beat analysis of non-invasive finger photoplethysmography (PPG) in AF patients restoring sinus rhythm by electrical cardioversion (ECV). Continuous non-invasive arterial blood pressure and left middle finger PPG pulse oximetry waveform (POW) signals were continuously recorded before and after elective ECV of consecutive AF or atrial flutter (AFL) patients. The main metrics (mean, standard deviation, coefficient of variation), as well as a beat-to-beat analysis of the pulse pressure (PP) and POW beat-averaged value (aPOW), were computed to compare pre- and post-ECV phases. 53 patients (mean age 69 ± 8 years, 79% males) were enrolled; cardioversion was successful in restoring SR in 51 (96%) and signal post-processing was feasible in 46 (87%) patients. In front of a non-significant difference in mean PP (pre-ECV: 51.96 ± 13.25, post-ECV: 49.58 ± 10.41 mmHg; p = 0.45), mean aPOW significantly increased after SR restoration (pre-ECV: 0.39 ± 0.09, post-ECV: 0.44 ± 0.06 a.u.; p < 0.001). Moreover, at beat-to-beat analysis linear regression yielded significantly different slope (m) for the PP (RR) relationship compared to aPOW(RR) [PP(RR): 0.43 ± 0.18; aPOW(RR): 1.06 ± 0.17; p < 0.001]. Long (> 95th percentile) and short (< 5th percentile) RR intervals were significantly more irregular in the pre-ECV phases for both PP and aPOW; however, aPOW signal suffered more fluctuations compared to PP (p < 0.001 in both phases). Present findings suggest that AF-related hemodynamic alterations are more manifest at the peripheral (aPOW) rather than at the upstream macrocirculatory level (PP). Restoring sinus rhythm increases mean peripheral microvascular perfusion and decreases variability of the microvascular hemodynamic signals. Future dedicated studies are required to determine if AF-induced peripheral microvascular alterations might relate to long-term prognostic effects.

Photoplethysmography waveform analysis for classification of vascular tone and arterial blood pressure: Study based on neural networks

To test whether a Shallow Neural Network (S-NN) can detect and classify vascular tone dependent changes in arterial blood pressure (ABP) by advanced photopletysmographic (PPG) waveform analysis. PPG and invasive ABP signals were recorded in 26 patients undergoing scheduled general surgery. We studied the occurrence of episodes of hypertension (systolic arterial pressure (SAP) >140 mmHg), normotension and hypotension (SAP < 90 mmHg). Vascular tone according to PPG was classified in two ways: 1) By visual inspection of changes in PPG waveform amplitude and dichrotic notch position; where Classes I-II represent vasoconstriction (notch placed >50% of PPG amplitude in small amplitude waves), Class III normal vascular tone (notch placed between 20-50% of PPG amplitude in normal waves) and Classes IV-V-VI vasodilation (notch <20% of PPG amplitude in large waves). 2) By an automated analysis, using S-NN trained and validated system that combines seven PPG derived parameters. The visual assessment was precise in detecting hypotension (sensitivity 91%, specificity 86% and accuracy 88%) and hypertension (sensitivity 93%, specificity 88% and accuracy 90%). Normotension presented as a visual Class III (III-III) (median and 1st-3rd quartiles), hypotension as a Class V (IV-VI) and hypertension as a Class II (I-III); all p < .0001. The automated S-NN performed well in classifying ABP conditions. The percentage of data with correct classification by S-ANN was 83% for normotension, 94% for hypotension, and 90% for hypertension. Changes in ABP were correctly classified automatically by S-NN analysis of the PPG waveform contour.

Blood pressure stratification using photoplethysmography and light gradient boosting machine.

Introduction: Globally, hypertension (HT) is a substantial risk factor for cardiovascular disease and mortality; hence, rapid identification and treatment of HT is crucial. In this study, we tested the light gradient boosting machine (LightGBM) machine learning method for blood pressure stratification based on photoplethysmography (PPG), which is used in most wearable devices. Methods: We used 121 records of PPG and arterial blood pressure (ABP) signals from the Medical Information Mart for Intensive Care III public database. PPG, velocity plethysmography, and acceleration plethysmography were used to estimate blood pressure; the ABP signals were used to determine the blood pressure stratification categories. Seven feature sets were established and used to train the Optuna-tuned LightGBM model. Three trials compared normotension (NT) vs. prehypertension (PHT), NT vs. HT, and NT + PHT vs. HT. Results: The F1 scores for these three classification trials were 90.18%, 97.51%, and 92.77%, respectively. The results showed that combining multiple features from PPG and its derivative led to a more accurate classification of HT classes than using features from only the PPG signal. Discussion: The proposed method showed high accuracy in stratifying HT risks, providing a noninvasive, rapid, and robust method for the early detection of HT, with promising applications in the field of wearable cuffless blood pressure measurement.

Respiratory patterns and baroreflex function in heart failure

Little is known on the effects of respiratory patterns on baroreflex function in heart failure (HF). Patients with HF (n = 30, age 61.6 ± 10 years, mean ± SD) and healthy controls (CNT, n = 10, age 58.9 ± 5.6 years) having their R–R interval (RRI, EKG), systolic arterial blood pressure (SBP, Finapres) and respiratory signal (RSP, Respitrace) monitored, were subjected to three recording sessions: free-breathing, fast- (≥ 12 bpm) and slow- (6 bpm) paced breathing. Baroreflex sensitivity (BRS) and power spectra of RRI, SBP, and RSP signals were calculated. During free-breathing, compared to CNT, HF patients showed a significantly greater modulation of respiratory volumes in the very-low-frequency (< 0.04 Hz) range and their BRS was not significantly different from that of CNT. During fast-paced breathing, when very-low-frequency modulations of respiration were reduced, BRS of HF patients was significantly lower than that of CNT and lower than during free breathing. During slow-paced breathing, BRS became again significantly higher than during fast breathing. In conclusion: (1) in free-breathing HF patients is present a greater modulation of respiratory volumes in the very-low-frequency range; (2) in HF patients modulation of respiration in the very-low and low frequency (around 0.1 Hz) ranges contributes to preserve baroreflex-mediated control of heart rate.

An Iterative Warping and Clustering Algorithm to Estimate Multiple Wave-Shape Functions From a Nonstationary Oscillatory Signal

Nonsinusoidal oscillatory signals are everywhere. In practice, the nonsinusoidal oscillatory pattern, modeled as a 1-periodic wave-shape function (WSF), might vary from cycle to cycle. When there are finite different WSFs, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$s_{1},\ldots,s_{K}$</tex-math></inline-formula> , so that the WSF jumps from one to another suddenly, the different WSFs and jumps encode useful information. We present an iterative warping and clustering algorithm to estimate <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$s_{1},\ldots,s_{K}$</tex-math></inline-formula> from a nonstationary oscillatory signal with time-varying amplitude and frequency, and hence the change points of the WSFs. The algorithm is a novel combination of time-frequency analysis, singular value decomposition entropy and vector spectral clustering. We demonstrate the efficiency of the proposed algorithm with simulated and real signals, including the voice signal, arterial blood pressure, electrocardiogram and accelerometer signal. Moreover, we provide a mathematical justification of the algorithm under the assumption that the amplitude and frequency of the signal are slowly time-varying and there are finite change points that model sudden changes from one wave-shape function to another one.

Classification of Blood Pressure Levels Based on Photoplethysmogram and Electrocardiogram Signals with a Concatenated Convolutional Neural Network.

Hypertension is a severe public health issue worldwide that significantly increases the risk of cardiac vascular disease, stroke, brain hemorrhage, and renal dysfunction. Early screening of blood pressure (BP) levels is essential to prevent the dangerous complication associated with hypertension as the leading cause of death. Recent studies have focused on employing photoplethysmograms (PPG) with machine learning to classify BP levels. However, several studies claimed that electrocardiograms (ECG) also strongly correlate with blood pressure. Therefore, we proposed a concatenated convolutional neural network which integrated the features extracted from PPG and ECG signals. This study used the MIMIC III dataset, which provided PPG, ECG, and arterial blood pressure (ABP) signals. A total of 14,298 signal segments were obtained from 221 patients, which were divided into 9150 signals of train data, 2288 signals of validation data, and 2860 signals of test data. In the training process, five-fold cross-validation was applied to select the best model with the highest classification performance. The proposed concatenated CNN architecture using PPG and ECG obtained the highest test accuracy of 94.56-95.15% with a 95% confidence interval in classifying BP levels into hypotension, normotension, prehypertension, hypertension stage 1, and hypertension stage 2. The result shows that the proposed method is a promising solution to categorize BP levels effectively, assisting medical personnel in making a clinical diagnosis.

Systolic Blood Pressure Estimation from PPG Signal Using ANN

High blood pressure is one of the most important precursors for Cardiovascular Diseases (CVDs), the most common cause of death in 2020, as reported by the World Health Organization (WHO). Moreover, many patients affected by neurodegenerative diseases (e.g., Parkinson’s Disease) exhibit impaired autonomic control, with inversion of the normal circadian arterial pressure cycle, and consequent augmented cardiovascular and fall risk. For all these reasons, a continuous pressure monitoring of these patients could represent a significant prognostic factor, and help adjusting their therapy. However, the existing cuff-based methods cannot provide continuous blood pressure readings. Our work is inspired by the newest approaches based on the photoplethysmographic (PPG) signal only, which has been used to continuously estimate systolic blood pressure (SP), using artificial neural networks (ANN), in order to create more compact and wearable devices. Our first database was derived from the PhysioNet resource; we extracted PPG and arterial blood pressure (ABP) signals, collected at a sampling frequency of 125 Hz, in a hospital environment. It consists of 249,672 PPG periods and the relative SP values. The second database was collected at STMicroelectronics s.r.l., in Agrate Brianza, using the MORFEA3 wearable device and a digital cuff-based sphygmomanometer, as reference. The pre-processing phase, in order to remove noise and motion artifacts and to segment the signal into periods, was carried out on Matlab R2019b. The noise removal was one of the challenging parts of the study because of the inaccuracy of the PPG signal during everyday-life activity, and this is the reason why the MORFEA3 dataset was acquired in a controlled environment in a static position. Different solutions were implemented to choose the input features that best represent the period morphology. The first database was used to train the multilayer feed-forward neural network with a back-propagation model, whereas the second one was used to test it. The results obtained in this project are promising and match the Association for the Advancement of Medical Instruments (AAMI) and the British Hypertension Society (BHS) standards. They show a Mean Absolute Error of 3.85 mmHg with a Standard Deviation of 4.29 mmHg, under the AAMI standard, and reach the grade A under the BHS standard.

Development and External Validation of a Machine Learning Model for the Early Prediction of Doses of Harmful Intracranial Pressure in Patients with Severe Traumatic Brain Injury.

Treatment and prevention of elevated intracranial pressure (ICP) is crucial in patients with severe traumatic brain injury (TBI). Elevated ICP is associated with secondary brain injury, and both intensity and duration of an episode of intracranial hypertension, often referred to as "ICP dose," are associated with worse outcomes. Prediction of such harmful episodes of ICP dose could allow for a more proactive and preventive management of TBI, with potential implications on patients' outcomes. The goal of this study was to develop and validate a machine-learning (ML) model to predict potentially harmful ICP doses in patients with severe TBI. The prediction target was defined based on previous studies and included a broad range of doses of elevated ICP that have been associated with poor long-term neurological outcomes. The ML models were used, with minute-by-minute ICP and mean arterial blood pressure signals as inputs. Harmful ICP episodes were predicted with a 30 min forewarning. Models were developed in a multi-center dataset of 290 adult patients with severe TBI and externally validated on 264 patients from the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) dataset. The external validation of the prediction model on the CENTER-TBI dataset demonstrated good discrimination and calibration (area under the curve: 0.94, accuracy: 0.89, precision: 0.87, sensitivity: 0.78, specificity: 0.94, calibration-in-the-large: 0.03, calibration slope: 0.93). The proposed prediction model provides accurate and timely predictions of harmful doses of ICP on the development and external validation dataset. A future interventional study is needed to assess whether early intervention on the basis of ICP dose predictions will result in improved outcomes.

Subject-Based Model for Reconstructing Arterial Blood Pressure from Photoplethysmogram.

The continuous prediction of arterial blood pressure (ABP) waveforms via non-invasive methods is of great significance for the prevention and treatment of cardiovascular disease. Photoplethysmography (PPG) can be used to reconstruct ABP signals due to having the same excitation source and high signal similarity. The existing methods of reconstructing ABP signals from PPG only focus on the similarities between systolic, diastolic, and mean arterial pressures without evaluating their global similarity. This paper proposes a deep learning model with a W-Net architecture to reconstruct ABP signals from PPG. The W-Net consists of two concatenated U-Net architectures, the first acting as an encoder and the second as a decoder to reconstruct ABP from PPG. Five hundred records of different lengths were used for training and testing. The experimental results yielded high values for the similarity measures between the reconstructed ABP signals and their reference ABP signals: the Pearson correlation, root mean square error, and normalized dynamic time warping distance were 0.995, 2.236 mmHg, and 0.612 mmHg on average, respectively. The mean absolute errors of the SBP and DBP were 2.602 mmHg and 1.450 mmHg on average, respectively. Therefore, the model can reconstruct ABP signals that are highly similar to the reference ABP signals.