Conference Paper

# A SVM Method for Continuous Blood Pressure Estimation from a PPG Signal

Authors:
If you want to read the PDF, try requesting it from the authors.

## Abstract

There is not always a linear relationship between the blood pressure and the pulse duration obtained from photoplethysmography (PPG) signal. In order to estimate the blood pressure from the PPG signal, A Support Vector Machine (SVM) method for continuous blood pressure estimation from a PPG Signal is applied in this paper. Training data were extracted from The University of Queensland Vital Signs Dataset for better representation of possible pulse and pressure variation. In total there were more than 7000 heartbeats and 9 parameters to be extracted from each other for analysis, then these features were defined as the input vector for training. The comparison between estimated and reference values shows better accuracy than the linear regression method and also shows better accuracy than the ANN method in diastolic blood pressure, which brings great significance in the field of mobile wearable.

## No full-text available

... Key features such as amplitudes and cardiac part phases were extracted through a fast Fourier transformation (FFT) and used to train an artificial neural network (ANN), which was then used to estimate BP using PPG. In [27], the support vector machine (SVM) algorithm showed better accuracy than the linear regression method and ANN. ...
... Therefore, 107 features encompassing seventy-five t-domain, sixteen f-domain, and ten statistical features were derived for each PPG signal along with six demographic data. The t-domain, f-domain, and statistical features were identified from different previous works [3,4,9,23,[25][26][27]38,39]. It is reported in the literature that 1-24 and 42-58 features were used in PPG related works [49]. ...
... Cattivelli et al. [25] introduced an algorithm for estimating BP, but used a very small amount of data (34 recordings for 25 subjects). Zhang et al. [27] described the SVM and neural network approach using time-domain features, which is used directly for the study of BP regression, and good results were obtained compared to the previous work. In [59], Zadi et al. showed the calculation of systolic and diastolic BP from PPG measurements using a viable method for continuous and noninvasive measurement of BP, however, using a very small dataset (15 subjects only). ...
Preprint
Full-text available
Hypertension is a potentially unsafe health ailment, which can be indicated directly from the Blood pressure (BP). Hypertension always leads to other health complications. Continuous monitoring of BP is very important; however, cuff-based BP measurements are discrete and uncomfortable to the user. To address this need, a cuff-less, continuous and a non-invasive BP measurement system is proposed using Photoplethysmogram (PPG) signal and demographic features using machine learning (ML) algorithms. PPG signals were acquired from 219 subjects, which undergo pre-processing and feature extraction steps. Time, frequency and time-frequency domain features were extracted from the PPG and their derivative signals. Feature selection techniques were used to reduce the computational complexity and to decrease the chance of over-fitting the ML algorithms. The features were then used to train and evaluate ML algorithms. The best regression models were selected for Systolic BP (SBP) and Diastolic BP (DBP) estimation individually. Gaussian Process Regression (GPR) along with ReliefF feature selection algorithm outperforms other algorithms in estimating SBP and DBP with a root-mean-square error (RMSE) of 6.74 and 3.59 respectively. This ML model can be implemented in hardware systems to continuously monitor BP and avoid any critical health conditions due to sudden changes.
... Key features such as amplitudes and cardiac part phases were extracted through a fast Fourier transformation (FFT) and used to train an artificial neural network (ANN), which was then used to estimate BP using PPG. In [27], the support vector machine (SVM) algorithm showed better accuracy than the linear regression method and ANN. ...
... Therefore, 107 features encompassing seventy-five t-domain, sixteen f-domain, and ten statistical features were derived for each PPG signal along with six demographic data. The t-domain, f-domain, and statistical features were identified from different previous works [3,4,9,23,[25][26][27]38,39]. It is reported in the literature that 1-24 and 42-58 features were used in PPG related works [49]. ...
... Cattivelli et al. [25] introduced an algorithm for estimating BP, but used a very small amount of data (34 recordings for 25 subjects). Zhang et al. [27] described the SVM and neural network approach using time-domain features, which is used directly for the study of BP regression, and good results were obtained compared to the previous work. In [59], Zadi et al. showed the calculation of systolic and diastolic BP from PPG measurements using a viable method for continuous and noninvasive measurement of BP, however, using a very small dataset (15 subjects only). ...
Article
Full-text available
Hypertension is a potentially unsafe health ailment, which can be indicated directly from the blood pressure (BP). Hypertension always leads to other health complications. Continuous monitoring of BP is very important; however, cuff-based BP measurements are discrete and uncomfortable to the user. To address this need, a cuff-less, continuous, and noninvasive BP measurement system is proposed using the photoplethysmograph (PPG) signal and demographic features using machine learning (ML) algorithms. PPG signals were acquired from 219 subjects, which undergo preprocessing and feature extraction steps. Time, frequency, and time-frequency domain features were extracted from the PPG and their derivative signals. Feature selection techniques were used to reduce the computational complexity and to decrease the chance of over-fitting the ML algorithms. The features were then used to train and evaluate ML algorithms. The best regression models were selected for systolic BP (SBP) and diastolic BP (DBP) estimation individually. Gaussian process regression (GPR) along with the ReliefF feature selection algorithm outperforms other algorithms in estimating SBP and DBP with a root mean square error (RMSE) of 6.74 and 3.59, respectively. This ML model can be implemented in hardware systems to continuously monitor BP and avoid any critical health conditions due to sudden changes.
... They used the support vector regression (SVR) technique as their estimator. Zhang and Feng [15] utilized the time intervals of the PPG as the features and reduced them to nine principal components. They employed an SVR to estimate the BP. ...
... We compared our method with state-of-the-art BP estimation algorithms [4,7,15,16] using the MAE, SDE, SDAE, and RMSD metrics ( Table 2). The methods in [4,7,16] used the Physionet dataset, while in [15] a non-public dataset was used. ...
... We compared our method with state-of-the-art BP estimation algorithms [4,7,15,16] using the MAE, SDE, SDAE, and RMSD metrics ( Table 2). The methods in [4,7,16] used the Physionet dataset, while in [15] a non-public dataset was used. As is shown in Table 2, the proposed method is more ...
Article
Full-text available
Accurate and uninterrupted estimation of the blood pressure is essential for continuous monitoring of patients. We estimate the blood pressure by extracting 21 time parameters from the photoplethysmography signal. The major novelties of this paper include: (1) using a nonlinear mapping to reduce the size of the feature vector and to map the input parameters to a latent space instead of conventional dimensionality reduction schemes, (2) employing a multi-stage noise reduction technique to effectively smooth the input signal. Estimation of the blood pressures is performed by a support vector regressor. The mean absolute errors of our results are 1.21 mmHg and 0.80 mmHg for systolic and diastolic blood pressures, respectively, which are lower than recent researches.
... Liu et al. reported 8.54 (MD) ± 10.9 (SDD) for SBP estimation using PPG-derived features and support vector machine (SVM) on the MIMIC dataset collected from patients in intensive care units [23]. Another study by Zhang et al. showed 11.6 ± 8.2 mmHg for SBP estimation using PPGderived features and SVM on the University of Queensland Vital Signs dataset collected from patients undergoing surgery [24]. Ruiz-Rodríguez et al. detailed a study where they achieved a performance of 2.98 ± 19.35 mmHg on patients in the critical care department and postanesthesia care unit using deep learning technology [25]. ...
... The SD of difference was within the acceptable threshold of 8 mmHg according to the AAMI standard [22]. The achieved performance was also superior compared to existing works [7,23,24] in the mean and SD of difference and compared to the work by Xing and et al. [26] in the SD of difference. ...
... While computational complexity has been rarely reported in existing studies, the application of advanced machine learning or deep learning models can result in high computational complexity. In this study, as the proposed method included dynamic feature selection and a linear regression with 9 data points at each time of recalibration, the computational complexity of the proposed method is O(N), lower than existing studies using SVM and deep learning models [23][24][25]. ...
Article
Full-text available
Background and Objectives In a significant portion of surgeries, blood pressure (BP) is often measured non-invasively in an intermittent manner. This practice has a risk of missing clinically relevant BP changes between two adjacent intermittent BP measurements. This study proposes a method to non-invasively estimate systolic blood pressure (SBP) with high accuracy in patients undergoing surgery. Methods Continuous arterial BP, electrocardiography (ECG), and photoplethysmography (PPG) signals were acquired from 29 patients undergoing surgery. After extracting 9 features from the PPG and ECG signals, we dynamically selected features upon each intermittent measurement (every 10 min) of SBP based on feature robustness and the principle of correlation-based feature selection. Finally, multiple linear regression models were built to combine these features to estimate SBP every 30 s. Results Compared to the reference SBP, the proposed method achieved a mean of difference at 0.08 mmHg, a standard deviation of difference at 7.97 mmHg, and a correlation coefficient at 0.89 (p < 0.001). Conclusions This study demonstrates the feasibility of non-invasively estimating SBP every 30 s with high accuracy during surgery by using ECG, PPG, and intermittent SBP measurements every 10 min, which meets the standard of the Association for the Advancement of Medical Instrumentation. The proposed method has the potential to enhance BP monitoring in the operating room, improving patient outcomes and experiences.
... However, they are restricted to limited scenarios (i.e., works in daytime and night, respectively) and their performance is always influenced by many factors (e.g., season effects, time of day, and temperature). Besides, machine learning algorithms are also adopted to improve the accuracy by extracting features from ECG and PPG [21,47,58]. However, collecting sufficient training data is inconvenient and usually takes a lot of effort. ...
... Extracting features from PPG, ECG [21,47,58] and using machine learning methods usually achieve enhanced accuracy and more consistent output. Nevertheless, these approaches have deficiencies in model training. ...
... In order to further evaluate the impact of HRV parameters on the accuracy of BP prediction, we compare this work with related research in terms of data sets, methods and estimation errors in table 5. Liu et al (Liu et al 2017) combined the second derivative of PPG (SDPPG) features with the conventional PPG features to train SVR-based BP estimators, but the data set is small and has been strictly selected through discarding signals whose APG is not 'W' shaped. Zhang et al (Zhang and Feng 2017) simplified the number of PPG feature parameters from 21 to 9, and the SVM algorithm improved the accuracy of BP prediction and reduced the complexity. Similarly, the amount of data they used was small, and only waves with insignificant dicrotic pulses were selected for universality. ...
... Thambiraj et al (Thambiraj et al 2020) considered the Womersley number along with different ECG and pulse features in the time domain and evaluated the performance of various ML methods. It should be noted that in the related studies detailed above, the hold-out validation was mostly used for model validation (Kachuee et al 2015, Zhang and Feng 2017, Lee et al 2019, Thambiraj et al 2020. However, the static 'hold-out method' is more sensitive to the partitioning method of the data set, which may lead to different models. ...
Article
Full-text available
Objective.Noninvasive blood pressure (BP) measurement technologies have been widely studied, but they still have the disadvantages of low accuracy, the requirement for frequent calibration and limited subjects. This work considers the regulation of vascular activity by the sympathetic nervous system and proposes a method for estimating BP using multiple physiological parameters.Approach.The parameters used in the model consist of heart rate variability (HRV), pulse transit time (PTT) and pulse wave morphology features extracted from electrocardiogram (ECG) and photoplethysmogram (PPG) signals. Through four classic machine learning algorithms, a hybrid data set of 3337 subjects from two databases is evaluated to verify the ability of cross-database migration. We also recommend an individual calibration procedure to further improve the accuracy of the method.Main results.The mean absolute error (MAE) and the root mean square error (RMSE) of the proposed algorithm is 10.03 and 14.55 mmHg for systolic BP (SBP), and 5.42 and 8.19 mmHg for diastolic BP (DBP). With individual calibration, the MAE and standard deviation (SD) is -0.16 ± 7.96 (SBP) and -0.13 ± 4.50 (DBP) mmHg, which satisfied the Advancement of Medical Instrumentation (AAMI) standard. In addition, the models are used to test single databases to evaluate their performance on different data sources. The overall performance of the Adaboost algorithm is better on the Multi-parameter Intelligent Monitoring in Intensive Care Unit (MIMIC) database; the MAE between its predicted value and true value reaches 6.6mmHg (SBP) and 3.12mmHg (DBP), respectively.Significance.The proposed method considers the regulation of blood vessels and the heart by the autonomic nervous system, and verifies its effectiveness and robustness across data sources, which is promising for improving the accuracy of continuous and cuffless BP estimation.
... As another way to predict BP without a cuff, some researchers have attempted to predict BP using a single raw biomedical signal or transform rather than the PWV-based feature [9], [10], [11], [12]. The concept of predicting BP using a single signal measurement such as ECG or PPG appears to be more appropriate for mobile devices such as smartphones and smart watches. ...
... The accuracy of BP prediction was found to be SBP 4.37 and DBP 3.95. Zhang [12] also extracted features from a PPG signal and performed a study to predict BP using the SVM method, which achieved SBP 11.64 ± 8.20 and DBP 7.62 ± 6.78. Tanveer [15] proposed a waveform-based hierarchical artificial neural network-long short-term memory (ANN-LSTM) model for BP estimation that automatically learns features from ECG and PPG signals through ANN and then uses them as inputs to LSTM to predict BP. ...
Article
Full-text available
Cardiovascular disease is the leading cause of death in the world. It is vital to prevent it by rapid diagnosis and appropriate management through periodic blood pressure (BP) measurement. Recently, many studies have been conducted on methods to measure BP without a cuff. One of the most common methods of predicting BP without a cuff is to use the correlation between pulse wave velocity (PWV) and BP. Studies that predict BP through PWV have two problems to overcome: 1) Additional efforts are required to extract PWV features manually from various biomedical signals, such as electrocardiogram (ECG) and photoplethysmogram (PPG); and 2) in predicting BP using biomedical signals from other people, individual periodic calibration is required because the correlation between PWV and BP differs from person to person. In this study, we proposed a cuffless BP prediction method based on a deep convolutional neural network (CNN) that can overcome the problems mentioned above. The proposed CNN method 1) can use raw signals for training without PWV feature extraction; and 2) automatically learns the characteristics of biomedical signals from other people to predict BP accurately without calibration. We propose two schemes: extraction through multiple dilated convolution, and concentration through strided convolution with a large kernel, to process sequential ECG and PPG signals through CNN. BP prediction performance was the best when both ECG and PPG signals were used together. To this end, we conducted extensive experiments on the different settings of the proposed method and constructed an effective learning model. The proposed method achieved excellent performance in predicting both systolic blood pressure and diastolic blood pressure over other known approaches. We also verified that the performance of our method fulfills international standard protocols, AAMI, and BHS.
... SBP and DBP values were extracted from the ABP signal by detecting the systolic and diastolic peaks as shown in Fig. 2. During our study, we selected 3 highly cited regression algorithms (SVM, AdaBoost, Random Forest) [9]- [11], [13], [14] and introduced 2 new boosting based regression algorithms -XGBoost [12], and CatBoost [18], [19]) in the domain of cuff-less BP estimation. Python's 'Scikit-learn library' is used for the first three algorithms [9]. ...
... The accuracy achieved by this algorithm in estimating cuffless BP is mentioned in Table II. This algorithm achieved better accuracy (MAE) than highly cited existing algorithms [9]- [14]. The Pearson's correlation coefficient value for the Catboost algorithm between target BP values and predicted BP values for the three BP categories (SBP, DBP, MAP) is found to be 0.37, 0.55, and 0.61. ...
Conference Paper
Full-text available
Blood Pressure (BP) is a critical biomarker for cardiorespiratory health. Conventional non-invasive BP measurement devices are mostly built on the principle of aus-cultation, oscillometry, or tonometry. The strong correlation between the Pulse Arrival Time (PAT) and BP has enabled unconstrained cuff-less BP monitoring. In this paper, we exploited that relationship for estimating Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), and Mean Arterial pressure (MAP) values. The proposed model involves extraction of PAT values by denoising the signals using advanced filtering techniques and finally employing machine learning algorithms to estimate cuff-less BP. The results are validated against Advancement of Medical Instrumentation (AAMI) standards and British Hypertension Society (BHS) protocols. The proposed method meets the AAMI standards in the context of estimating DBP and MAP values. The model's accuracy achieved Grade A for both MAP and DBP values using the CatBoost algorithm, whereas it achieved grade A for MAP and Grade B for DBP using the XGBoost algorithm based on the BHS standards.
... The findings indicate that modeling the BP dynamics' temporal dependencies increases the long-term BP prediction accuracy dramatically [47]. In addition, the authors in [48] investigated, with fair results, the likelihood of utilizing raw PPG data to detect arrhythmias using wearable devices in real-time, demonstrating the possibility of using the raw PPG signal as inputs for deep learners (RNN-LSTM). Their approach achieved a receiver operator characteristic (ROC) curve of 0.9999, with false positive and false negative rates both below 2 × 10 −3 . ...
Article
Full-text available
High blood pressure (BP) may lead to further health complications if not monitored and controlled, especially for critically ill patients. Particularly, there are two types of blood pressure monitoring, invasive measurement, whereby a central line is inserted into the patient's body, which is associated with infection risks. The second measurement is cuff-based that monitors BP by detecting the blood volume change at the skin surface using a pulse oximeter or wearable devices such as a smartwatch. This paper aims to estimate the blood pressure using machine learning from photo-plethysmogram (PPG) signals, which is obtained from cuff-based monitoring. To avoid the issues associated with machine learning such as improperly choosing the classifiers and/or not selecting the best features, this paper utilized the tree-based pipeline optimization tool (TPOT) to automate the machine learning pipeline to select the best regression models for estimating both systolic BP (SBP) and diastolic BP (DBP) separately. As a pre-processing stage, notch filter, band-pass filter, and zero phase filtering were applied by TPOT to eliminate any potential noise inherent in the signal. Then, the automated feature selection was performed to select the best features to estimate the BP, including SBP and DBP features, which are extracted using random forest (RF) and k-nearest neighbors (KNN), respectively. To train and test the model, the PhysioNet global dataset was used, which contains 32.061 million samples for 1000 subjects. Finally, the proposed approach was evaluated and validated using the mean absolute error (MAE). The results obtained were 6.52 mmHg for SBS and 4.19 mmHg for DBP, which show the superiority of the proposed model over the related works.
... In SVR, instead of trying to minimize the error outright as with traditional regression techniques, hyperplanes are constructed to fit the error observed within a small threshold of ε (epsilon). Previous studies have used support vector-based techniques on PPG data to predict other continuous biological variables such as blood pressure estimation [43]. ...
Article
Full-text available
Diabetes is a chronic illness that affects millions of people worldwide and requires regular monitoring of a patient’s blood glucose level. Currently, blood glucose is monitored by a minimally invasive process where a small droplet of blood is extracted and passed to a glucometer—however, this process is uncomfortable for the patient. In this paper, a smartphone video-based noninvasive technique is proposed for the quantitative estimation of glucose levels in the blood. The videos are collected steadily from the tip of the subject’s finger using smartphone cameras and subsequently converted into a Photoplethysmography (PPG) signal. A Gaussian filter is applied on top of the Asymmetric Least Square (ALS) method to remove high-frequency noise, optical noise, and motion interference from the raw PPG signal. These preprocessed signals are then used for extracting signal features such as systolic and diastolic peaks, the time differences between consecutive peaks (DelT), first derivative, and second derivative peaks. Finally, the features are fed into Principal Component Regression (PCR), Partial Least Square Regression (PLS), Support Vector Regression (SVR) and Random Forest Regression (RFR) models for the prediction of glucose level. Out of the four statistical learning techniques used, the PLS model, when applied to an unbiased dataset, has the lowest standard error of prediction (SEP) at 17.02 mg/dL.
... Several time-related and amplitude-related features are proposed in the state-of-the-art [22], [23]. The extracted features are then mapped to blood pressure values using different techniques; such as, multiple linear regression (MLR) [24], [25], artificial neural networks (ANN) [26], [27], support vector machine (SVM) [28], random forests (RF) [24], etc. It is also worth mentioning that with the increasing interest in deep learning techniques, recent works have suggested PWA techniques using raw PPG signals as input to deep neural networks (DNN), without the need for explicit feature extraction [29]. ...
Preprint
In this work, we present the Senbiosys blood pressure monitoring algorithm (SB-BPM) that solely requires a photoplethysmography (PPG) signal. The technology is based on pulse wave analysis (PWA) of PPG signals retrieved from different body locations to continuously estimate the systolic blood pressure (SBP) and the diastolic blood pressure (DBP).
... Figure 2 illustrates the prediction of our model and the ground truth graphically. These results are comparable to that of other current works such as that of Zhang et al. [40] or Hasanzadeh et al. [34] (see Section 5 for a detailed analysis on the performance). ...
Article
Full-text available
Blood pressure (BP) is among the most important vital signals. Estimation of absolute BP solely using photoplethysmography (PPG) has gained immense attention over the last years. Available works differ in terms of used features as well as classifiers and bear large differences in their results. This work aims to provide a machine learning method for absolute BP estimation, its interpretation using computational methods and its critical appraisal in face of the current literature. We used data from three different sources including 273 subjects and 259,986 single beats. We extracted multiple features from PPG signals and its derivatives. BP was estimated by xgboost regression. For interpretation we used Shapley additive values (SHAP). Absolute systolic BP estimation using a strict separation of subjects yielded a mean absolute error of 9.456mmHg and correlation of 0.730. The results markedly improve if data separation is changed (MAE: 6.366mmHg, r: 0.874). Interpretation by means of SHAP revealed four features from PPG, its derivation and its decomposition to be most relevant. The presented approach depicts a general way to interpret multivariate prediction algorithms and reveals certain features to be valuable for absolute BP estimation. Our work underlines the considerable impact of data selection and of training/testing separation, which must be considered in detail when algorithms are to be compared. In order to make our work traceable, we have made all methods available to the public.
... (1) Accurate measurement of PTT requires at least a single channel ECG and a single or dual-channel PPG acquisition modules. 17 (2) Compared to PPG, the acquisition of ECG is quite complicated. ...
Article
Full-text available
Recently, photoplethysmography (PPG)-based techniques have been extensively used for cuff-less, automated estimation of blood pressure because of their inexpensive and effortless acquisition technology compared to other conventional approaches. However, most of the reported PPG-based, generalized BP estimation methods often lack the desired accuracy due to pathophysiological diversity. Moreover, some methods rely on several correction factors, which are not globalized yet and require further investigation. In this paper, a simple and automated systolic (SBP) and diastolic (DBP) blood pressure estimation method is proposed based on patient-specific neural network (NN) modeling. Initially, 15 time-plane PPG features are extracted and after feature selection, only four selected features are used in the NN model for beat-to-beat estimation of SBP and DBP, respectively. The proposed technique also presents reasonable accuracy while used for generalized estimation of BP. Performance of the algorithm is evaluated on 670 records of 50 intensive care unit (ICU) patients taken from MIMIC, MIMIC II and MIMIC Challenge databases. The proposed algorithm exhibits high average accuracy with (mean[Formula: see text][Formula: see text][Formula: see text]SD) of the estimated SBP as ([Formula: see text]) mmHg and DBP as ([Formula: see text]) mmHg. Compared to the other generalized models, the use of patient-specific approach eliminates the necessity of individual correction factors, thus increasing the robustness, accuracy and potential of the method to be implemented in personal healthcare applications.
... Since 2016, studies on blood pressure estimation [9,11,12], biometric identification [13,14], and atrial fibrillation detection [15][16][17][18] from PPG signals using deep learning has become popular. Some of these studies were able to use readily available public databases for training the deep learning models [19][20][21], but others required conducting large-scale experiments to produce the necessary data [15,22,23]. Because deep learning models need large datasets for proper optimization, the availability of data in large quantities becomes a limiting factor in developing such methods, as generating properly annotated databases require vast resources which can be afforded by a few. ...
Article
Full-text available
Photoplethysmogram (PPG) is one of the most widely measured biosignals alongside electrocardiogram (ECG). Due to the simplicity of measurement and the advent of wearable devices, there have been growing interest in using PPG for a variety of healthcare applications such as cardiac function estimation. However, unlike ECG, there are not many large databases available for clinically significant analyses of PPG. To overcome this issue, a Generative Adversarial Network-based model to generate PPG using ECG as input is proposed. The network was trained using a large open database of biosignals measured from surgical patients and was externally validated using an alternative database sourced from another hospital. The generated PPG was compared with the reference PPG using percent root mean square difference (PRD) and Pearson correlation coefficient to evaluate the morphological similarity. Additionally, heart rate measured from the reference ECG, reference PPG, and generated PPG, and compared through repeated measure analysis of variance to test for any significant differences. The mean PRD was 32±10% and the mean correlation coefficient was 0.95±0.05 in the test dataset. The HR from the three biosignals showed no significant difference with a p-value of 0.473. When the optimized GAN model was tested on atrial fibrillation ECG from a third dataset, the mean correlation coefficient between the generated PPG heart rate and the ECG heart rate was 0.94±0.15, with paired t-test resulting in p-value of 0.64. The results indicate that the proposed method may provide a valuable alternative to augmenting biosignal databases that are abundant in one signal while lacking in another.
... Since, these studies have used different datasets and/or smaller number of subjects for evaluation than the proposed methodology, therefore, fair comparison can't be performed for BP estimation. The studies performed by [30]- [31] have utilized the University of Queensland Vital Signs Database (32 surgical persons) and their own datasets [32], [34] respectively, wherein our proposed methodology performed better compared to these studies despite of having much larger and diverse population with CVD complications. ...
Article
This paper presents a deep learning model ‘PP-Net’ which is the first of its kind, having the capability to estimate the physiological parameters: Diastolic blood pressure (DBP), Systolic blood pressure (SBP), and Heart rate (HR) simultaneously from the same network using a single channel PPG signal. The proposed model is designed by exploiting the deep learning framework of Long-term Recurrent Convolutional Network (LRCN), exhibiting inherent ability of feature extraction, thereby, eliminating the cost effective steps of feature selection and extraction, making less-complex for deployment on resource constrained platforms such as mobile platforms. The performance demonstration of the PP-Net is done on a larger and publically available MIMIC-II database. We achieved an average NMAE of 0.09 (DBP) and 0.04 (SBP) mmHg for BP, and 0.046 bpm for HR estimation on total population of 1557 critically ill subjects. The accurate estimation of HR and BP on a larger population compared to the existing methods, demonstrated the effectiveness of our proposed deep learning framework. The accurate evaluation on a huge population with CVD complications, validates the robustness of the proposed framework in pervasive healthcare monitoring especially cardiac and stroke rehabilitation monitoring.
... Machine learning algorithms have been used in bioengineering for pathologies detection and classification. The support vector machines (SVM) has been used in the diagnosis of diabetes [11], the detection of epileptic episodes [12], the classification of gut microbiomes [13], the estimation of blood pressure [14] and the classification of Alzheimer [15]. More specifically, SVM has also been used in the classification of cardiac diseases [16], cardiac abnormalities [17] and arrhythmia classification [18][19][20]. ...
... Zhang e Feng apresentam um método para estimativa da pressão sistólica e diastólica com base na medição de parâmetros correspondentes ao formato do pulso do sinal PPG, e aplicam este conjunto de parâmetros ao treinamento de uma rede neural(Zhang and Feng, 2017). Também utilizam o método SVM (Support Vector Machine) para modelar e análisar os dados e estimar a pressão sistólica e diastólica de um conjunto de pacientes, efetuando a validação dos dados por medição direta da pressão nestes pacientes. ...
... With the development of machine learning, the characteristic parameters of models were further enriched, including the amplitude, phase characteristics of pulse waves extracted with fast Fourier transform [23], spectral characteristics [24], and the features of the photoplethysmography (PPG) waveform and related first and second (time) derivatives [25,26]. Moreover, the model construction methods were expanded, such as neural network [24,[27][28][29], support vector machine [30], adaptive boosting regression [31], and random forest algorithm [32]. The blood-pressure estimation methods based on machine learning and big data covered more blood-pressure information and improved the estimation accuracies of the models to some extent. ...
Article
Full-text available
Blood pressure is an extremely important blood hemodynamic parameter. The pulse wave contains abundant blood-pressure information, and the convenience and non-invasivity of its measurement make it ideal for non-invasive continuous monitoring of blood pressure. Based on combined photoplethysmography and electrocardiogram signals, this study aimed to extract the waveform information, introduce individual characteristics, and construct systolic and diastolic blood-pressure (SBP and DBP) estimation models using the back-propagation error (BP) neural network. During the model construction process, the mean impact value method was employed to investigate the impact of each feature on the model output and reduce feature redundancy. Moreover, the multiple population genetic algorithm was applied to optimize the BP neural network and determine the initial weights and threshold of the network. Finally, the models were integrated for further optimization to generate the final individualized continuous blood-pressure monitoring models. The results showed that the predicted values of the model in this study correlated significantly with the measured values of the electronic sphygmomanometer. The estimation errors of the model met the Association for the Advancement of Medical Instrumentation (AAMI) criteria (the SBP error was 2.5909 ± 3.4148 mmHg, and the DBP error was 2.6890 ± 3.3117 mmHg) and the Grade A British Hypertension Society criteria.
... Multiparameter fusion approaches based on machine learning algorithms (e.g. support vector machines, random forests and neural networks) were proposed to improve the PTT-based models by merging various parameters extracted from electrocardiogram (ECG) and photoplethysmography (PPG) influence BP and showed better performance [9]- [11]. However, features should be manually extracted in such approaches and thus deep features related to BP changes have not been really excavated. ...
Conference Paper
Full-text available
Cuff-less blood pressure (BP) monitoring is increasingly being needed for cardiovascular events management in clinical. Many of the existing methods, however, are based on manual feature extraction, which cannot characterize the complex relationship between the physiological signals and BP. In this study, the 16-layer VGGNet was used to construct cuff-less BP from electrocardiogram (ECG) and pressure pulse wave (PPW) signals, with no need extract features from raw signals. The deep network architecture has the ability of automatic feature learning, and the learned features are the higher-level abstract description of low-level raw physiological signals. Eight-nine middle-aged and elderly subjects were enrolled to evaluate the performance of the proposed BP estimation method, with oscillometric technique-based BP as a reference. Experimental results indicate that the proposed method had a commendable accuracy in BP estimation, with a correlation coefficient of 0.91 and an estimation error of -2.06 ± 6.89 mmHg for systolic BP, and 0.89 and -4.66 ± 4.91 mmHg for diastolic BP. This study shows that the proposed method provided a potential novel insight for the cuff-less BP estimation.
... This is probably due to the reason that the features are carefully selected based on physiological reasons. Zhang and Feng [49] in 2017 extracted many detailed features from a PPG signal and performed a study to predict BP using the SVM method, which achieved SBP: 11.64 ± 8.20 and DBP: 7.62 ± 6.78. ...
Article
Recent works on the machine learnings designed for cuffless blood pressure (BP) estimation based on measured photoplethysmogram (PPG) waveforms are reviewed by this study, with future trends of the related technology developments distilled. This review starts with those based on the conventional pulse wave velocity (PWV) theory, by which few equations are derived to calculate BPs based on measured pulse arrival times (PATs) and/or pulse transit time (PTT). Due to the inadequacy of PATs and PPTs to characterize BP, some works were reported to employ more features in PPG waveforms to achieve better accuracy. In these works, varied machine learnings were adopted, such as support vector machine (SVM), regression tree (RT), adaptive boosting (AdaBoost), and artificial neural network (ANN), etc., resulting in satisfactory accuracies based on a large number of data in the databases of Queensland and/or MIMIC II. Most recently, a few studies reported to utilize the deep learning machines like convolution neural network (CNN), recursive neural network (RNN), and long short-term memory (LSTM), etc., to handle feature extraction and establish models integrally, with the aim to cope with the inadequacy of pre-determined (hand-crafted) features to characterize BP and the difficulty of extracting pre-determined features by a designed algorithm. Therefore, the deep learning opens an opportunity of achieving much better BP accuracy by using a single PPG sensor. Favorable accuracies have been resulted by these few studies in comparison with prior works. Finally, future research efforts needed towards successful commercialization of the cuffless BP sensor are distilled.
... In our future work, we will focus on using a simpler and more effective sensor to collect PPW signals and to validate the accuracy of the PPW-based method for BP measurement further, with more participants to make the PPW-based model more reliable. Furthermore, some machine-learning techniques such as artificial neural networks [12], support vector machines [51], and random forests [52] have been reported to have the potential for BP measurement. Therefore, we will attempt to validate the application of machine-learning techniques in selecting optimal sets of PPW features, and then establish a PPW-based BP estimation model with higher accuracy. ...
Article
Full-text available
Pulse transit time (PTT) has received considerable attention for noninvasive cuffless blood pressure measurement. However, this approach is inconvenient to deploy in wearable devices because two sensors are required for collecting two-channel physiological signals, such as electrocardiogram and pulse wave signals. In this study, we investigated the pressure pulse wave (PPW) signals collected from one piezoelectric-induced sensor located at a single site for cuffless blood pressure estimation. Twenty-one features were extracted from PPW that collected from the radial artery, and then a linear regression method was used to develop blood pressure estimation models by using the extracted PPW features. Sixty-five middle-aged and elderly participants were recruited to evaluate the performance of the constructed blood pressure estimation models, with oscillometric technique-based blood pressure as a reference. The experimental results indicated that the mean ± standard deviation errors for the estimated systolic blood pressure and diastolic blood pressure were 0.70 ± 7.78 mmHg and 0.83 ± 5.45 mmHg, which achieved a decrease of 1.33 ± 0.37 mmHg in systolic blood pressure and 1.14 ± 0.20 mmHg in diastolic blood pressure, compared with the conventional PTT-based method. The proposed model also demonstrated a high level of robustness in a maximum 60-day follow-up study. These results indicated that PPW obtained from the piezoelectric sensor has great feasibility for cuffless blood pressure estimation, and could serve as a promising method in home healthcare settings.
... Applying machine learning techniques also in a regression setting is of particular interest in this field as it enables new non-invasive monitoring techniques for several physiological signals, such as arterial blood pressure (ABP). Research has been conducted to estimate APB from several other signals, such as Photoplethysmogram (PPG) [3] or Electrocardiogram (ECG) and heart rate [4]. ...
Article
Full-text available
The biomedical field is characterized by an ever-increasing production of sequential data, which often come in the form of biosignals capturing the time-evolution of physiological processes, such as blood pressure and brain activity. This has motivated a large body of research dealing with the development of machine learning techniques for the predictive analysis of such biosignals. Unfortunately, in high-stakes decision making, such as clinical diagnosis, the opacity of machine learning models becomes a crucial aspect to be addressed in order to increase the trust and adoption of AI technology. In this paper, we propose a model agnostic explanation method, based on occlusion, that enables the learning of the input’s influence on the model predictions. We specifically target problems involving the predictive analysis of time-series data and the models that are typically used to deal with data of such nature, i.e., recurrent neural networks. Our approach is able to provide two different kinds of explanations: one suitable for technical experts, who need to verify the quality and correctness of machine learning models, and one suited to physicians, who need to understand the rationale underlying the prediction to make aware decisions. A wide experimentation on different physiological data demonstrates the effectiveness of our approach both in classification and regression tasks.
... Most previous studies on machine-learning derived early warning algorithms are limited by relying only one a single data source such as arterial pressure waveforms or photoplethysmographs [13][14][15]. However, hemodynamic changes are also associated with alteration in physiological profiles, including ECG and EEG [16]. In addition, early warning models based on arterial blood pressure (ABP) waveforms necessarily require extensive feature engineering based on proprietary algorithms. ...
Article
Full-text available
To develop deep learning models for predicting Interoperative hypotension (IOH) using waveforms from arterial blood pressure (ABP), electrocardiogram (ECG), and electroencephalogram (EEG), and to determine whether combination ABP with EEG or CG improves model performance. Data were retrieved from VitalDB, a public data repository of vital signs taken during surgeries in 10 operating rooms at Seoul National University Hospital from January 6, 2005, to March 1, 2014. Retrospective data from 14,140 adult patients undergoing non-cardiac surgery with general anaesthesia were used. The predictive performances of models trained with different combinations of waveforms were evaluated and compared at time points at 3, 5, 10, 15 minutes before the event. The performance was calculated by area under the receiver operating characteristic (AUROC), area under the precision-recall curve (AUPRC), sensitivity and specificity. The model performance was better in the model using both ABP and EEG waveforms than in all other models at all time points (3, 5, 10, and 15 minutes before an event) Using high-fidelity ABP and EEG waveforms, the model predicted IOH with a AUROC and AUPRC of 0.935 [0.932 to 0.938] and 0.882 [0.876 to 0.887] at 5 minutes before an IOH event. The output of both ABP and EEG was more calibrated than that using other combinations or ABP alone. The results demonstrate that a predictive deep neural network can be trained using ABP, ECG, and EEG waveforms, and the combination of ABP and EEG improves model performance and calibration.
... Notable examples for portable, continuous non-invasive BP monitoring include (i) automated oscillometric methods, 66 (ii) ultrasonic device-based methods, 67 and (iii) softwarebased BP estimation methods. 68 Oscillometric methods which use cuffs are the most commonly used devices; the difficulty with them, however, is that (i) they are uncomfortable to wear and (ii) they have an inherent error with respect to their ability in representing the arterial BP values. 69 The feasibility in estimating BP from the SpO 2 and ECG signals has been previously explored with little success 70 and ultrasonic devices are yet to become mainstream in continuous BP monitoring. ...
Article
Full-text available
The pandemic has brought to everybody’s attention the apparent need of remote monitoring, highlighting hitherto unseen challenges in healthcare. Today, mobile monitoring and real-time data collection, processing and decision-making, can drastically improve the cardiorespiratory–haemodynamic health diagnosis and care, not only in the rural communities, but urban ones with limited healthcare access as well. Disparities in socioeconomic status and geographic variances resulting in regional inequity in access to healthcare delivery, and significant differences in mortality rates between rural and urban communities have been a growing concern. Evolution of wireless devices and smartphones has initiated a new era in medicine. Mobile health technologies have a promising role in equitable delivery of personalized medicine and are becoming essential components in the delivery of healthcare to patients with limited access to in-hospital services. Yet, the utility of portable health monitoring devices has been suboptimal due to the lack of user-friendly and computationally efficient physiological data collection and analysis platforms. We present a comprehensive review of the current cardiac, pulmonary, and haemodynamic telemonitoring technologies. We also propose a novel low-cost smartphone-based system capable of providing complete cardiorespiratory assessment using a single platform for arrhythmia prediction along with detection of underlying ischaemia and sleep apnoea; we believe this system holds significant potential in aiding the diagnosis and treatment of cardiorespiratory diseases, particularly in underserved populations.
... Several time-related and amplitude-related features are proposed in the state-of-the-art [22], [23]. The extracted features are then mapped to blood pressure values using different techniques, such as multiple linear regression (MLR) [24], [25], artificial neural networks (ANN) [26], [27], support vector machine (SVM) [28], random forests (RF) [24], etc. It is also worth mentioning that with the increasing interest in deep learning techniques, recent works have suggested PWA techniques using raw PPG signals as input to deep neural networks (DNN), without the need for explicit feature extraction [29]. ...
Article
Full-text available
In this work, we present a photoplethysmography-based blood pressure monitoring algorithm (PPG-BPM) that solely requires a photoplethysmography (PPG) signal. The technology is based on pulse wave analysis (PWA) of PPG signals retrieved from different body locations to continuously estimate the systolic blood pressure (SBP) and the diastolic blood pressure (DBP). The proposed algorithm extracts morphological features from the PPG signal and maps them to SBP and DBP values using a multiple linear regression (MLR) model. The performance of the algorithm is evaluated on the publicly available Multiparameter Intelligent Monitoring in Intensive Care (MIMIC I) database. We utilize $28$ data-sets (records) from the MIMIC I database that contain both PPG and brachial arterial blood pressure (ABP) signals. The collected PPG and ABP signals are synchronized and divided into intervals of $30$ seconds, called epochs. In total, we utilize $47153$ \textit{clean} $30$ -second epochs for the performance analysis. Out of the $28$ data-sets, we use only $2$ data-sets (records $041$ and $427$ in the MIMIC I) with a total of $2677$ \textit{clean} $30$ -second epochs to build the MLR model of the algorithm. For the SBP, a standard deviation of error (SDE) of $8.01$ mmHg and a mean absolute error (MAE) of $6.10$ mmHg between the arterial line and the PPG-based values are achieved, with a Pearson correlation coefficient $r = 0.90$ , $p<.001$ . For the DBP, an SDE of $6.22$ mmHg and an MAE of $4.65$ mmHg between the arterial line and the PPG-based values are achieved, with a Pearson correlation coefficient $r = 0.85$ , $p<.001$ . We also use a binary classifier for the BP values with the positives indicating SBP $\geq 130$ mmHg and/or DBP $\geq 80$ mmHg and the negatives indicating otherwise. The classifier results generated by the PPG-based SBP and DBP estimates achieve a sensitivity and a specificity of $79.11\%$ and $92.37\%$ , respectively.
... In addition to the use of artificial neural networks and other network models, traditional machine learning method [18,27,28] was also used in many studies to optimize the feature extraction process. ...
Preprint
Full-text available
Blood pressure indicates cardiac function and peripheral vascular resistance and is critical for disease diagnosis. Traditionally, blood pressure data are mainly acquired through contact sensors, which require high maintenance and may be inconvenient and unfriendly to some people (e.g., burn patients). In this paper, an efficient non-contact blood pressure measurement network based on face videos is proposed for the first time. An innovative oversampling training strategy is proposed to handle the unbalanced data distribution. The input video sequences are first normalized and converted to our proposed YUVT color space. Then, the Spatio-temporal slicer encodes it into a multi-domain Spatio-temporal mapping. Finally, the neural network computation module, used for high-dimensional feature extraction of the multi-domain spatial feature mapping, after which the extracted high-dimensional features are used to enhance the time-domain feature association using LSTM, is computed by the blood pressure classifier to obtain the blood pressure measurement intervals. Combining the output of feature extraction and the result after classification, the blood pressure calculator, calculates the blood pressure measurement values. The solution uses a blood pressure classifier to calculate blood pressure intervals, which can help the neural network distinguish between the high-dimensional features of different blood pressure intervals and alleviate the overfitting phenomenon. It can also locate the blood pressure intervals, correct the final blood pressure values and improve the network performance. Experimental results on two datasets show that the network outperforms existing state-of-the-art methods.
Conference Paper
Blood pressure (BP) is an important indicator for prevention and management of cardiovascular diseases. Alongside the improvement in sensors and wearables, photoplethysmography (PPG) appears to be a promising technology for continuous, non-invasive and cuffless BP monitoring. Previous attempts mainly focused on features extracted from the pulse morphology. In this paper, we propose to remove the feature engineering step and automatically generate features from an ensemble average (EA) PPG pulse and its derivatives, using convolutional neural network and a calibration measurement. We used the large VitalDB dataset to accurately evaluate the generalization capability of the proposed model. The model achieved mean errors of -0.24 ± 11.56 mmHg for SBP and -0.5 ± 6.52 mmHg for DBP. We observed a considerable reduction in error standard deviation of above 40% compared to the control case, which assumes no BP variation. Altogether, these results highlight the capability to model the dependency between PPG and BP.
Article
High blood pressure (BP) or hypertension is the single most crucial adjustable risk factor for cardiovascular diseases (CVDs) and monitoring the arterial blood pressure (ABP) is an efficient way to detect and control the prevalence of the cardiovascular health of patients. Therefore, monitoring the regulation of BP during patients’ daily life plays a critical role in the ambulatory setting and the latest mobile health technology. In recent years, many studies have been conducted to explore the feasibility and performance of such techniques in the health care system. The ultimate aim of these studies is to find and develop an alternative to conventional BP monitoring by using cuff-less, easy-to-use, fast, and cost-effective devices for controlling and lowering the physical harm of CVDs to the human body. However, most of the current studies are at the prototype phase and face a range of issues and challenges to meet clinical standards. This review focuses on the description and analysis of the latest continuous and cuff-less methods along with their key challenges and barriers. Particularly, most advanced and standard technologies including pulse transit time (PTT), ultrasound, pulse arrival time (PAT), and machine learning are investigated. The accuracy, portability, and comfort of use of these technologies, and the ability to integrate to the wearable healthcare system are discussed. Finally, the future directions for further study are suggested.
Article
Given that current cuffless blood pressure (BP) measurement technologies feature acceptable overall accuracy, this paper proposed a sufficiently accurate cuffless BP estimation method based on photoplethysmography (PPG) and electrocardiography (ECG) signals. This study used single-channel PPG and ECG signals to estimate heart rate (HR), diastolic BP (DBP), and systolic BP (SBP). A modified long-term recurrent convolutional network comprising a multi-scale convolution network and a long short-term memory (LSTM) network was used to develop a deep learning model for accurately estimating BP and HR. The PPG and ECG signal data of 1551 patients were obtained from the Data Sets-UCI Machine Learning Repository of the University of California, Irvine. The study dataset comprised ECG, PPG, and arterial BP (ABP) signals from the PhysioNet MIMIC II dataset. The original signals were processed by removing noise and artifacts. The aforementioned dataset contains 12,000 records in a hierarchical data format, with each record containing three signals, namely 125-Hz ECG signals from channel II (ECG lead II), 125-Hz PPG signals from the fingertip, and 125-Hz invasive ABP signals. To validate the stability and performance of the developed model, ten-fold cross-validation was conducted. The mean absolute error (MAE) (standard deviation (SD)) values of the developed model for predicting SBP, DBP, and HR were 2.24 mmHg (3.59 mmHg), 1.40 mmHg (2.56 mmHg), and 0.84 bpm (2.23 bpm), respectively. In addition, the estimated SBP and DBP values satisfied the standards of the British Hypertension Society and the Association for the Advancement of Medical Instrumentation. Compared with the methods proposed in other studies, the deep learning model developed in this study required a lower number of layers to provide accurate SBP, DBP, and HR estimations. The results of this study confirmed the effectiveness of the proposed deep learning architecture.
Article
Arterial blood pressure (ABP) monitoring may permit the early diagnosis and management of cardiovascular disease (CVD); however, existing methods for measuring ABP outside the clinic use inconvenient cuff sphygmomanometry, or do not estimate continuous ABP waveforms. This study proposes a novel deep learning model DeepCNAP for estimating continuous BP waveform from a noninvasively measured photoplethysmography (PPG) signal in real time. DeepCNAP was designed through the combination of deep convolutional networks and self-attention. The proposed method was constructed via 10-fold cross-validation based on the MIMIC database (the number of subjects = 942, recording time = 374.43 hours). The performance of DeepCNAP was evaluated from two perspectives: estimating ABP from PPG and classifying hemodynamically unstable events (i.e., hypertension, prehypertension, hypotension, and the normal state). The mean absolute errors of the BP estimates were 3.40 4.36 mmHg for systolic BP, 1.75 2.25 mmHg for diastolic BP, and 3.23 2.21 mmHg for the BP waveform, indicating that DeepCNAP satisfies the standards of both the British Hypertension Society (BHS) and the Association for the Advancement of Medical Instrumentation (AAMI). From the estimated BP, hypertension, prehypertension, hypotension, and the normal state were classified with 99.44, 97.58, 92.23, and 94.64% accuracy, respectively. DeepCNAP enables feasible real-time estimation of invasively measured ABP from noninvasive PPG. With its noninvasive nature, high accuracy, and clinical relevance, the proposed model could be valuable in general wards at hospitals and for wearable devices in daily life.
Article
Full-text available
Arterial blood pressure is not only an important index that must be measured in routine physical examination but also a key monitoring parameter of the cardiovascular system in cardiac surgery, drug testing, and intensive care. To improve the measurement accuracy of continuous blood pressure, this paper uses photoplethysmography (PPG) signals to estimate diastolic blood pressure and systolic blood pressure based on ensemble empirical mode decomposition (EEMD) and temporal convolutional network (TCN). In this method, the clean PPG signal is decomposed by EEMD to obtain n-order intrinsic mode functions (IMF), and then the IMF and the original PPG are input into the constructed TCN neural network model, and the results are output. The results show that TCN has better performance than CNN, CNN-LSTM, and CNN-GRU. Using the data added with IMF, the results of the above neural network model are better than those of the model with only PPG as input, in which the systolic blood pressure (SBP) and diastolic blood pressure (DBP) results of EEMD-TCN are −1.55 ± 9.92 mmHg and 0.41 ± 4.86 mmHg. According to the estimation results, DBP meets the requirements of the AAMI standard, BHS evaluates it as Grade A, SD of SBP is close to the standard AAMI, and BHS evaluates it as Grade B.
Conference Paper
Heart disease and stroke are the leading causes of death worldwide. High blood pressure greatly increases the risk of heart disease and stroke. Therefore, it is important to control blood pressure (BP) through regular BP monitoring; as such, it is necessary to develop a method to accurately and conveniently predict BP in a variety of settings. In this paper, we propose a method for predicting BP without feature extraction using fully convolutional neural networks (CNNs). We measured single multi-wave photoplethysmography (PPG) signals using a smartphone. To find an effective wavelength of PPG signals for the generation of accurate BP measurements, we investigated the BP prediction performance by changing the combinations of the input PPG signals. Our CNN-based BP predictor yielded the best performance metrics when a green PPG time signal was used in combination with an instantaneous frequency signal. This combination had an overall mean absolute error (MAE) of 5.28 and 4.92 mmHg for systolic and diastolic BP, respectively. Thus, our CNN-based approach achieved comparable results to other approaches that use a single PPG signal.
Conference Paper
Blood Pressure (BP) is one of the four primary vital signs indicating the status of the body's vital (life-sustaining) functions. BP is difficult to continuously monitor using a sphygmomanometer (i.e. a blood pressure cuff), especially in everyday-setting. However, other health signals which can be easily and continuously acquired, such as photoplethysmography (PPG), show some similarities with the Aortic Pressure waveform. Based on these similarities, in recent years several methods were proposed to predict BP from the PPG signal. Building on these results, we propose an advanced personalized data-driven approach that uses a three-layer deep neural network to estimate BP based on PPG signals. Different from previous work, the proposed model analyzes the PPG signal in time-domain and automatically extracts the most critical features for this specific application, then uses a variation of recurrent neural networks (RNN) called Long-Short-Term-Memory (LSTM) to map the extracted features to the BP value associated with that time window. Experimental results on two separate standard hospital datasets, yielded absolute errors mean and absolute error standard deviation for systolic and diastolic BP values outperforming prior works.
Preprint
Blood Pressure (BP) is one of the four primary vital signs indicating the status of the body's vital (life-sustaining) functions. BP is difficult to continuously monitor using a sphygmomanometer (i.e. a blood pressure cuff), especially in everyday-setting. However, other health signals which can be easily and continuously acquired, such as photoplethysmography (PPG), show some similarities with the Aortic Pressure waveform. Based on these similarities, in recent years several methods were proposed to predict BP from the PPG signal. Building on these results, we propose an advanced personalized data-driven approach that uses a three-layer deep neural network to estimate BP based on PPG signals. Different from previous work, the proposed model analyzes the PPG signal in time-domain and automatically extracts the most critical features for this specific application, then uses a variation of recurrent neural networks called Long-Short-Term-Memory (LSTM) to map the extracted features to the BP value associated with that time window. Experimental results on two separate standard hospital datasets, yielded absolute errors mean and absolute error standard deviation for systolic and diastolic BP values outperforming prior works.
Article
Full-text available
A new neural network (NN) is orchestrated by this study to achieve high-accuracy in blood pressure (BP) estimation by a real-time photoplethysmography (PPG). The PPG system consists of an OLED/OPD module to detect the pulsation of blood vessels, followed by a readout circuitry. The circuit is comprised of transimpedance amplifier, a digital tune high order band pass filter, programmable gain amplifier (PGA), time interleave OLED driver, micro-controller unit, and the Bluetooth transceiver. The obtained PPG signals are subsequently processed with quality checking, feature extraction, and into an NN for estimating BP. The feature extraction is assisted, by principal component analysis (PCA) to reduce the total number of input features to five with accuracy assured. 96 subjects participated in data collection for calibrating the designed NN. The resulted correlation is 0.81, while the errors for SBP and DBP are 2.00 ± 6.08 and 1.87 ± 4.09 mmHg, respectively. According to the Advancement of Medical Instrumentation (AAMI) and British Hypertension Society (BHS), a BP device in Grade A needs to control its accuracy error less than ± 8 mmHg, based on which the BP sensor developed herein are in Grade A, since the resulted errors of ± 6.08 and ± 4.09 mmHg are both less than ± 8 mmHg, showing the satisfactory performance of the BP monitor developed by this study.
Article
In this paper, we propose a continuous physiological parameter estimation model based on a deep learning network for photoplethysmography (PPG) sensor signals. Signals of 8-s duration were incorporated into the proposed model in this study for frequent estimation of the systolic blood pressure (BP), diastolic BP, heart rate (HR), and mean arterial pressure of the human body; this facilitated early identification and monitoring of physiological conditions and thus reduced the risk of cardiovascular disease. The proposed model was designed using a convolutional neural network (CNN) and long short-term memory (LSTM) network. This model was trained and validated using the large-scale Multiparameter Intelligent Monitoring in Intensive Care database. The CNN was used to extract features from PPG signals automatically. This automatic extraction replaced the conventional manual feature extraction process. Features with time-series were then analyzed using the LSTM network to estimate physiological parameters. Subsequently, ten-fold cross-validation was conducted to reveal the mean absolute errors ± standard deviations of participants’ systolic BP, diastolic BP, HR, and mean arterial pressure to be 2.54 ± 3.88, 1.59 ± 2.45, 1.62 ± 2.55, and 1.59 ± 2.34 mmHg, respectively. These values meet the standards established by the Association for the Advancement of Medical Instrumentation and the British Hypertension Society. The proposed method facilitates the accurate, continuous monitoring of the BP and HR.
Article
At present, the mainstream blood glucose detection methods are invasive, which will cause harm to the human body and make it inconvenient to measure. The non-contact measurement method can avoid these problems. In this paper, a non-contact blood glucose detection method based on a near-infrared camera is proposed. Blood glucose has a strong absorption capacity in the near-infrared band, and other components in blood (water, hemoglobin, etc.) have different absorption characteristics in this band compared with blood glucose. Therefore, in this method, we realize blood glucose detection by receiving the near-infrared light reflected back after blood glucose absorption. We extracted 26 pulse wave features from the pulse wave and analyzed 6 that were highly correlated with blood glucose. Then, four kinds of machine learning algorithms (PCR, PLS, SVR, RFR) were used to build models respectively, and the RFR with the best performance was selected to build the final blood glucose prediction model. Finally, the experimental results are analyzed by Clark error grid analysis, which shows that the proposed method is in good agreement with the reference glucose monitor. Compared with traditional invasive blood glucose detection methods, the non-contact blood glucose detection method has more application prospects.
Chapter
The field of mobile health (mHealth) focuses on the uses of mobile technologies to support the delivery of healthcare services and management of health in everyday life. mHealth tools range from clinical applications for remote patient monitoring and shared decision making to tools intended to help individuals better manage chronic conditions or make health-promoting lifestyle changes. In addition, mHealth tools leverage sensors, phone-based questionnaires, and activity inference to facilitate low-burden data collection in research, greatly enhancing our ability to understand how context, activity, and physiological processes interact to shape health and health behavior. This chapter traces the history of mHealth, highlights key features of these technologies that make them uniquely suited for delivery of health interventions, and overviews areas of health services and health promotion for which they are used. The chapter concludes with a brief discussion of ethical issues raised by the rapid growth of mHealth.
Article
The machine learning approach has gained a significant attention in the healthcare sector because of the prospect of developing new techniques for medical devices and handling the critical database of chronic diseases. The learning approach has potential to analyze complex medical data, disease diagnosis, and patient monitoring system, and to monitor e-health record. Non-invasive cuffless blood pressure (CLBP) measurement secured a significant position in the patient monitoring system. From a few recent decades, the importance of cuffless technology has been perceived towards continuous monitoring of blood pressure (BP) and supplementary efforts have been made towards its continuous monitoring. However, the optimal method that measures BP unambiguously and continuously has not yet emerged along with issues like calibration time, accuracy and long-term estimation of BP with miniaturizing hardware. The present study provides an insight into several learning algorithms along with their feature selection models. Various challenges and future improvements towards the current state of machine learning in healthcare industries are discussed in the present review. The bottom line of this study is to provide a comprehensive perspective of the machine learning approach of CLBP for the generation of highly precise predictive models for continuous BP measurement.
Article
Blood pressure monitoring is very important for the prevention of cardiovascular diseases. In this paper, we proposed a multi-type features fusion (MTFF) neural network model for blood pressure (BP) prediction based on photoplethysmography (PPG). The model includes two convolutional neural networks (CNN) which used to train the morphological and frequency spectrum features of PPG signal, and one Bi-directional long short term memory (BLSTM) network which used to train the temporal features of PPG signal. These multi-features were fused through a specific fusion module after training, so more information of PPG signals were obtained and the hidden relationship between the fused features and blood pressure was established. The standard deviation (STD) and mean absolute error (MAE) of the fusion model are 7.25 mmHg and 5.59 mmHg respectively for systolic blood pressure (SBP), 4.48 mmHg and 3.36 mmHg respectively for diastolic blood pressure (DBP). The results are in full compliance with the Advancement of Medical Instrumentation (AAMI) and the British Hypertension Society (BHS) international standards. We conclude that the MTFF neural network proposed in this paper can accurately predict blood pressure. The significant difference from the traditional methods of BP prediction based on manual calculation of features is that our method automatically extracts PPG features through the deep learning model which can easily handle the complicated and tedious calculation. Compared with other similar BP prediction methods based on deep learning, three different features are trained and fused, which further improves the accuracy of BP prediction.
Article
Full-text available
Goal: Continuous Blood Pressure monitoring can provide invaluable information about individuals' health conditions. However, BP is conventionally measured using inconvenient cuff-based instruments, which prevents continuous BP monitoring. This work presents an efficient algorithm, based on the Pulse Arrival Time (PAT), for the continuous and cuff-less estimation of the Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), and Mean Arterial Pressure (MAP) values. Methods: The proposed framework estimates the BP values through processing vital signals and extracting two types of features, which are based on either physiological parameters or whole-based representation of vital signals. Finally, the regression algorithms are employed for the BP estimation. Although the proposed algorithm works reliably without any need for calibration, an optional calibration procedure is also suggested, which can improve the system's accuracy even further. Results: The proposed method is evaluated on about a thousand subjects using the Association for the Advancement of Medical Instrumentation (AAMI) and the British Hypertension Society (BHS) standards. The method complies with the AAMI standard in the estimation of DBP and MAP values. Regarding the BHS protocol, the results achieve grade A for the estimation of DBP and grade B for the estimation of MAP. Conclusion: We conclude that by using the PAT in combination with informative features from the vital signals, the BP can be accurately and reliably estimated in a non-invasive fashion. Significance: The results indicate that the proposed algorithm for the cuff-less estimation of the blood pressure can potentially enable mobile health-care gadgets to monitor the BP continuously.
Article
Full-text available
This paper proposes a new calibration parameter from the dicrotic notch of photoplethysmography (PPG) waveform for systolic blood pressure (SBP) estimation using pulse transit time (PTT). An experiment including exercise was conducted on twelve subjects. The results show that estimation with PTT and a new parameter, i.e. the Relative Amplitude of Secondary peak (RAS), can predict SBP within 1.7 plusmn 6.8 mmHg of the reference for all the subjects before and after exercise (including within 5 minutes and 20-40 minutes after exercise). The underestimated bias after exercise if PTT was used alone can be largely reduced when RAS is introduced. The preliminary analysis indicates that RAS is a promising parameter for calibrating the PTT-based approach for cuffless BP estimation.
Chapter
Full-text available
Smartphones have become one of the widest and often used devices that people bring almost every time and everywhere. Their computational capacities allow their application to many every-day tasks. One of them is health state monitoring. This chapter presents a smartphone-based photoplethysmogram (PPG) acquisition and pulse rate evaluation system. The proposal was designed for different smartphone models, equipped with a LED or not. Different cameras represent the same acquired information in different ways: changes may occur in color saturation, resolution, frame rate, etc. Therefore, several smartphones were used to define the common characteristics of the captured video, and establish proper criteria for PPG extraction. Moreover, the appropriate algorithms were proposed and validated to verify the correct device usage, the system calibration, the PPG and pulse rate evaluation. The experimental results have confirmed the correctness and suitability of the proposed method with respect to the medical pulse measurement instruments.
Chapter
Full-text available
Blood pressure is often measured using a device called a sphygmomanometer, a stethoscope, and a blood pressure cuff. All the existing manual or automatic measuring techniques of blood pressure are based on this principle, which is not convenient for continuous monitoring of blood pressure. In this paper, we proposed the regression model which could estimate unspecified people’s systolic blood pressure (SBP) conveniently and continuously and checked its accuracy with blood pressure cuff. The method for estimating each individual SBP by using only pulse wave transit time (PWTT) has been studied, but it is difficult to estimate unspecified people’s SBP with the method using only PWTT. This study examines the relationships between arterial blood pressure and certain features of the photoplethysmographic (PPG) signals from 10 healthy subjects. The experiment involved three sessions, which is the resting period, exercise period and recovery period.
Conference Paper
Full-text available
We study the problem of noninvasively estimating Blood Pressure (BP) without using a cuff, which is attractive for continuous monitoring of BP over Body Area Networks. It has been shown that the Pulse Arrival Time (PAT) measured as the delay between the ECG peak and a point in the finger PPG waveform can be used to estimate systolic and diastolic BP. Our aim is to evaluate the performance of such a method using the available MIMIC database, while at the same time improve the performance of existing techniques. We propose an algorithm to estimate BP from a combination of PAT and heart rate, showing improvement over PAT alone. We also show how the method achieves recalibration using an RLS adaptive algorithm. Finally, we address the use case of ECG and PPG sensors wirelessly communicating to an aggregator and study the effect of skew and jitter on BP estimation.
Article
Full-text available
In order to detect hypertension at its early stage, blood pressure (BP) of an individual must be carefully monitored. Aiming to monitor BP noninvasively and continuously, this study examines a new feature, normalized harmonic area (NHA), which is extracted from photoplethysmoographic (PPG) signals in the period domain by using the discrete period transform (DPT). BP, pulse transit time (PTT), diastolic time (DT) and NHA were obtained from photoplethysmographic (PPG) signals in the period domain by using the discrete period transform (DPT). BP, pulse transit time (PTI), diastolic time (DT) and NHA were obtained from 28 subjects before and immediately after step-climbing exercise. It was found that NHA has more significant correlation wiith BP than PTT and DT in this study. The mean difference and standard deviation (Mean±SD) between the BP estimated from NHA and the reference BP obtained from a commercial oscillometric BP meter were 0.37±4.3 mmHg and 0.51±4.8 mmHg for SBP and DBP, respectively. The result indicate that NHA is potentially a useful indicator of arterial BP.
The increasing availability of low cost and easy to use personalized medical monitoring devices has opened the door for new and innovative methods of health monitoring to emerge. Cuff-less and continuous methods of measuring blood pressure are particularly attractive as blood pressure is one of the most important measurements of long term cardiovascular health. Current methods of noninvasive blood pressure measurement are based on inflation and deflation of a cuff with some effects on arteries where blood pressure is being measured. This inflation can also cause patient discomfort and alter the measurement results. In this work, a mobile application was developed to collate the PhotoPlethysmoGramm (PPG) waveform provided by a pulse oximeter and the electrocardiogram (ECG) for calculating the pulse transit time. This information is then indirectly related to the user's systolic blood pressure. The developed application successfully connects to the PPG and ECG monitoring devices using Bluetooth wireless connection and stores the data onto an online server. The pulse transit time is estimated in real time and the user's systolic blood pressure can be estimated after the system has been calibrated. The synchronization between the two devices was found to pose a challenge to this method of continuous blood pressure monitoring. However, the implemented continuous blood pressure monitoring system effectively serves as a proof of concept. This combined with the massive benefits that an accurate and robust continuous blood pressure monitoring system would provide indicates that it is certainly worthwhile to further develop this system.
Article
LIBSVM is a library for Support Vector Machines (SVMs). We have been actively developing this package since the year 2000. The goal is to help users to easily apply SVM to their applications. LIBSVM has gained wide popularity in machine learning and many other areas. In this article, we present all implementation details of LIBSVM. Issues such as solving SVM optimization problems theoretical convergence multiclass classification probability estimates and parameter selection are discussed in detail.
Article
Simple and non-invasive methods to estimate vital signs are very important for preventive healthcare. In this paper, we present a methodology to estimate Blood Pressure (BP) using Photoplethysmography (PPG). Instead of directly relating systolic and diastolic BP values with PPG features, our proposed methodology initially maps PPG features with some person specific intermediate latent parameters and later derives BP values from them. The 2-Element Windkessel model has been considered in the current context to estimate total peripheral resistance and arterial compliance of a person using PPG features, followed by linear regression for simulating arterial blood pressure. Experimental results, performed on a standard hospital dataset yielded absolute errors of 0.78±13.1 mmHg and 0.59 ± 10.23 mmHg for systolic and diastolic BP values respectively. Results also indicate that the methodology is more robust than the standard methodologies that directly estimate BP values from PPG signal.
Conference Paper
This paper presents a new transformation technique called the Pulse Active transform (PAT). The PAT uses a series of harmonically related periodic triangular waveforms to decompose a signal into a finite set of pulse active features. These features incorporate the signal's information in the pulse active domain, and which are subsequently processed for some desired application. PAT is non-invertible thus ensuring complete security of the original signal source. In this paper PAT is demonstrated on an ECG signal and used for biometric authentication. The new transformation technique is tested on 112 PTB subjects. It is shown in this paper that the new transformation has a superior performance compared to the conventional characteristic based feature extraction methods with additional security to avoid recovery of the original ECG.
Article
In this paper, we propose a methodology to estimate the range of human blood pressure (BP) using Photoplethysmography (PPG). 12 time domain features and 7 frequency domain features are pointed out and extracted from the PPG signal. A feature selection algorithm based on Maximal Information Coefficient (MIC) is presented to reduce the dimensionality of the feature set to effective ones, thereby cutting down resource requirements. Support Vector Machine (SVM) is used to classify the BP values into separate bins. The proposed methodology is validated and tested on a standard benchmark clean dataset as well as phone captured noisy dataset to justify its robustness and efficiency. Apart from a commending performance improvement, BP estimation is achieved with minimal features and processing, making the algorithm light weight for porting on smart phones.
Conference Paper
Regular ECG check up is a good practice for cardiac patients as well as elderly people. In this paper we propose a low cost methodology to coarsely estimate the range of some important parameters of ECG using Photoplethysmography (PPG). PPG is easy to measure (even with a smart phone) and strongly related to human cardio-vascular system. The proposed methodology extracts a set of time domain features from PPG signal. A statistical analysis is performed to select the most relevant set of PPG features for the ECG parameters. Training model for the ECG parameters are created based on those selected features. Both artificial neural network and support vector machine based supervised learning approach is used for performance comparison. Experimental results, performed on benchmark dataset shows that good accuracy in the estimation of ECG parameters can be achieved in our proposed methodology. Results also show that the overall performance improves in using feature selection technique rather than using all the PPG features for classification.
Conference Paper
There is a relation, not always linear, between the blood pressure and the pulse duration, obtained from photoplethysmography (PPG) signal. In order to estimate the blood pressure from the PPG signal, in this paper the Artificial Neural Networks (ANNs) are used. Training data were extracted from the Multiparameter Intelligent Monitoring in Intensive Care waveform database for better representation of possible pulse and pressure variation. In total there were analyzed more than 15000 heartbeats and 21 parameters were extracted from each of them that define the input vector for the ANN. The comparison between estimated and reference values shows better accuracy than the linear regression method and satisfy the American National Standards of the Association for the Advancement of Medical Instrumentation.
Article
Data recorded from the devices used to monitor a patient's vital signs are often used in the development of displays, alarms, and information systems, but high-resolution, multiple-parameter datasets of anesthesia monitoring data from patients during anesthesia are often difficult to obtain. Existing databases have typically been collected from patients in intensive care units. However, the physical state of intensive care patients is dissimilar to those undergoing surgery, more frequent and marked changes to cardiovascular and respiratory variables are seen in operating room patients, and additional and highly relevant information to anesthesia (e.g., end-tidal agent monitoring, etc.) is omitted from these intensive care databases. We collected a set of high-quality, high-resolution, multiple-parameter monitoring data suitable for anesthesia monitoring research. Vital signs data were recorded from patients undergoing anesthesia at the Royal Adelaide Hospital. Software was developed to capture, time synchronize, and interpolate vital signs data from Philips IntelliVue MP70 and MP30 patient monitors and Datex-Ohmeda Aestiva/5 anesthesia machines into 10 millisecond resolution samples. The recorded data were saved in a variety of accessible file formats. Monitoring data were recorded from 32 cases (25 general anesthetics, 3 spinal anesthetics, 4 sedations) ranging in duration from 13 minutes to 5 hours (median 105 min). Most cases included data from the electrocardiograph, pulse oximeter, capnograph, noninvasive arterial blood pressure monitor, airway flow, and pressure monitor and, in a few cases, the Y-piece spirometer, electroencephalogram monitor, and arterial blood pressure monitor. Recorded data were processed and saved into 4 file formats: (1) comma-separated values text files with full numerical and waveform data, (2) numerical parameters recorded in comma-separated values files at 1-second intervals, (3) graphical plots of all waveform data in a range of resolutions as Portable Network Graphics image files, and (4) graphical overview plots of numerical data for entire cases as Portable Network Graphics and Scalable Vector Graphics files. The complete dataset is freely available online via doi:102.100.100/6914 and has been listed in the Australian National Data Service Collections Registry. The present dataset provides clinical anesthesia monitoring data from entire surgical cases where patients underwent anesthesia, includes a wide range of vital signs variables that are commonly monitored during surgery, and is published in accessible, user-friendly file formats. The text and image file formats let researchers without engineering or computer science backgrounds easily access the data using standard spreadsheet and image browsing software. In future work, monitoring data should be collected from a wider range and larger number of cases, and software tools are needed to support searching and navigating the database.
Article
There are many studies about cuffless and continuous blood pressure estimation using pulse transit time (PTT). In this study, we proposed the modeling method which could estimate systolic BP (SBP) conveniently and indirectly using PTT and some biometric parameters. 45 people participated in this study and we measured PTT using photoplethysmography (PPG) and electrocardiogram (ECG) signals and biometric parameters such as weight, height, body mass index (BMI), length of arm and circumference of arm. Before modeling, we selected variables as predictors using statistical analysis. With these parameters, we compared artificial neural network (ANN) with multiple regressions as an estimating method of BP. We evaluated the mean differences and standard deviations between estimated value and reference value, acquired from a KEDA-approved device. The results showed that the ANN had better accuracy than the multiple regression. ANN's estimation satisfied AAMI standard as a BP device.
Article
A method for estimating pulse wave velocity (PWV) using circulatory waveform signals derived from multiple photoplethysmograph (PPG) sensors is described. The method employs two wearable in-line PPG sensors placed at a known distance from one another at the ulnar and digital artery. A technique for calibrating the measured pulse wave velocity to arterial blood pressure using hydrostatic pressure variation is presented. Additionally, a framework is described for estimating local arterial dynamics using PPG waveforms and multi-channel blind system ID. Initial results implementing the method on data derived from a human subject at different arterial pressures is presented. Results show that the method is capable of measuring the changes in arterial PWV that result from fluctuations in mean arterial pressure.
Conference Paper
Arterial blood pressure is one of the most important physiological parameters for the noninvasive diagnosis and monitoring of cardiovascular diseases. This study examines the relationships between arterial blood pressure and certain features of the photoplethysmographic (PPG) signals obtained from 15 young healthy subjects. Width of 1/2 pulse amplitude, width of 2/3 pulse amplitude, systolic upstroke time and diastolic time of the pulse were selected as features of the PPG signals. The experiment involved three sessions: rest, step-climbing exercise, and recovery from exercise. It was found that the diastolic time has higher correlation with systolic blood pressure (SBP) and diastolic blood pressure (DBF) than other features. The mean differences between the estimated and the measured blood pressure using a standard device are 0.21 mm Hg for SBP and 0.02 mm Hg for DBF. The corresponding standard deviations are 7.32 mm Hg for SBP and 4.39 mm Hg for DBF. The preliminary results indicate that it is possible to use the photoplethysmography only for cuffless and continuous estimation of arterial blood pressure.
Conference Paper
Proposed paper is a part of a research to develop a convenient method for continuous monitoring of blood pressure by non-invasive and non-oscillometric way. The method is based on a presumption that there is a singular relationship between the velocity of pulse wave and blood pressure for a person. The detection of the velocity of pulse wave is based on the registration of the mechanical movements of blood vessel's walls and computing the time delay in different regions of the human body using an ECG as a reference signal. Additional information about the blood pressure evaluation is gotten from the photoplethysmographic (PPG) channel. Used optical method, which is able to detect pulsation profiles of the blood vessels, is based on the self-mixing that occurs in the diode laser cavity when the radiation, scattered back by the skin into the laser, interferes with the field inside it and causes the changes of the laser pump current. National Instruments software LabView for Windows is used for output signal digitalization and processing purposes.
A neural network-based method for continuous blood pressure estimation from a PPG signal
• Y Kurvylyak
• F Lamonaca
• D Grimaldi
LIBSVM: a library for support vector machines. A CM Transactions on Intelligent Systems and Technology
• C.-C Chang
• C.-J Lin
Signal and Data Processing for Measurement Systems
• Y Kurylyak
• F Lamonaca
• D Grimaldi