Cuffless Blood Pressure Estimation Research Articles

RU-Net, Multi-Scale Feature Fusion and Transfer Learning: Unlocking the Potential of Cuffless Blood Pressure Monitoring with PPG and ECG.

This study introduces an innovative deep-learning model for cuffless blood pressure estimation using PPG and ECG signals, demonstrating state-of-the-art performance on the largest clean dataset, PulseDB. The rU-Net architecture, a fusion of U-Net and ResNet, enhances both generalization and feature extraction accuracy. Accurate multi-scale feature capture is facilitated by short-time Fourier transform (STFT) time-frequency distributions and multi-head attention mechanisms, allowing data-driven feature selection. The inclusion of demographic parameters as supervisory information further elevates performance. On the calibration-based dataset, our model excels, achieving outstanding accuracy (SBP MAE ± std: 4.49 ± 4.86 mmHg, DBP MAE ± std: 2.69 ± 3.10 mmHg), surpassing AAMI standards and earning a BHS Grade A rating. Addressing the challenge of calibration-free data, we propose a fine-tuning-based transfer learning approach. Remarkably, with only 10% data transfer, our model attains exceptional accuracy (SBP MAE ± std: 4.14 ± 5.01 mmHg, DBP MAE ± std: 2.48 ± 2.93 mmHg). This study sets the stage for the development of highly accurate and reliable wearable cuffless blood pressure monitoring devices.

A paralleled CNN and Transformer network for PPG-based cuff-less blood pressure estimation

Cuff-less blood pressure measurement plays a pivotal part in hypertension diagnosis and prevention. By employing deep learning algorithms, blood pressure estimation applying photoplethysmography (PPG) signals has achieved remarkable improvements and attracted growing attention. However, most existing methods utilize the PPG signal emphasized on localized feature learning, often neglecting the global features in the temporal data. To tackle the issue, we propose a novel architecture. It leverages a parallel processing approach by using Convolutional Neural Network for extracting local features and simultaneously employing Transformer to capture global features. The feature fusion block integrates the features using spatial and channel attention. The blood pressure values are regressed by two fully connected layers. To verify the performance of our method, we conduct evaluations on the Medical Information Mart for Intensive Care (MIMIC) database. On the MIMIC database, our algorithm demonstrates exceptional performance. For Diastolic Pressure, Mean Arterial Pressure, and Systolic Pressure, the mean absolute errors are 2.36 mmHg, 2.03 mmHg, and 4.44 mmHg, respectively. This performance aligns with the rigorous criteria required by an Association for the Advancement of Medical Instrumentation (AAMI) and receives a Grade A rating in accordance with the British Hypertension Society (BHS) standard.

O18 UNCERTAINTY QUANTIFICATION OF WEARABLE CUFFLESS BLOOD PRESSURE MEASUREMENT

Background and Objective: Due to aleatoric and epistemic uncertainty, most cuffless blood pressure (BP) estimation models struggle to provide reliable and accurate BP estimations. The purpose of this study is to quantify the uncertainty of wearable cuffless BP estimation so as to reduce the impact of uncertainty on the accuracy of model estimation, and in the meanwhile to provide an estimated uncertainty interval (UI) in addition to the point estimation. Methods: We developed a gradient boosting regression tree (GBRT) model to estimate ambulatory BP and the estimation UI with eight noninvasive features extracted from wearable photoplethysmogram (PPG) and electrocardiogram (ECG) signals. We validated the proposed method with the Microsoft Aurora dataset that was originally collected for the Aurora-BP study. We identified 483 subjects (247 males) with wearable watches collecting PPG, ECG, and other signals, while ambulatory BP was monitored hourly using an oscillometric BP device. We trained the model with quantile loss on 60% of subjects (2954 samples), then calibrated the estimated UI with conformal predication with 24% (2148 samples) of subjects and tested the model with 16% (1658 samples) of the subjects. Results: The mean absolute difference (MAD) in systolic BP (SBP) and diastolic BP (DBP) estimated by the GBRT model were 13.96 mmHg and 9.89 mmHg, respectively. Then, with implementation of conformal prediction with error rates of ϵ=0.05, for the test set samples, the percentage of the estimated UI covering the reference BP during the daytime and nighttime phases can reach 91.67% and 95.89%, respectively. Conclusions: The combination of the GBRT model with conformal predication can quantify the uncertainty in cuffless BP estimations. In addition, the estimated uncertainty quantification interval together with the point estimation is potentially more reliable than single-point estimation, which could benefit better diagnosis and treatment of hypertension.

A new method to compute the blood flow equations using the physics-informed neural operator

Using the Navier-Stokes equation for real blood flow simulations in patients is a challenging task. Among many of the challenges, initial and boundary conditions are not directly measurable. Many parameters for the flow model are also not directly measurable and differ from patient to patient and over time. New machine learning techniques, like physics-informed neural networks (PINN), offer some new ways to handle these difficulties. In this work, we aim to learn the operator that maps some easily measurable physiological signals to the solution of the blood flow equation. We use our proposed model based on Navier-Stokes equation and PINN to fit real data on blood pressure. A Windkessel boundary condition is used to produce physically correct reflection waves. A time-periodic condition is used to capture the periodicity of blood flow and enables our model to simulate the blood flow without initial and boundary conditions. Furthermore, we allow the periods of each instance of solution to be different, which makes the training of neural operators computationally expensive, but more accuracy and physical correct towards real blood pressure data. Further more, we also propose an efficient implementation to incorporate the periodic condition into our model. Estimating the hyper-parameters in the Navier-Stokes equation is also difficult. We then introduce a hyper-parameter network to estimate these parameters during the training process as well. The blood flow data contains useful information for disease detection and diagnosis, but directly measuring the entire blood flow remains a significant challenge. We apply our proposed method to cuffless blood pressure estimation. More specifically, we aim to predict the blood pressure waveform (continuous blood pressure and velocity in both time and space) from Electrocardiogram (ECG) and photoplethysmogram (PPG) signals, which can be easily measured using wearable devices. Compared to other methods, our method is the first one that can predict blood flow continuously, both with location and time which are valuable for cardio-vascular medical treatments and diagnoses.

High-Detectivity All-Polymer Photodiode Empowers Smart Vitality Surveillance and Computational Imaging Rivaling Silicon Diodes.

Near-infrared (NIR) organic photodetectors (OPDs), particularly all-polymer-based ones, hold substantial commercial promise in the healthcare and imaging sectors. However, the process of optimizing their active layer composition to achieve highly competitive figures of merit lacks a clear direction and methodology. In this work, celebrity polymer acceptor PY-IT into a more NIR absorbing host system PBDB-T:PZF-V, to significantly enhance the photodetection competence, is introduced. The refined all-polymer ternary broadband photodetector demonstrates superior performance metrics, including experimentally measured noise current as low as 6 fA Hz-1/2, specific detectivity reaching 8×1012 Jones, linear dynamic range (LDR) of 145 dB, and swift response speed surpassing 200 kHz, striking a fair balance between sensitivity and response speed. Comprehensive morphological and photophysical characterizations elucidate the mechanisms behind the observed performance enhancements in this study, which include reduced trap density, enhanced charge transport, diminished charge recombination, and balanced electron/hole mobilities. Moreover, the practical deployment potential of the proof-of-concept device in self-powered mode is demonstrated through their application in a machine learning-based cuffless blood pressure (BP) estimation system and in high-resolution computational imaging across complex environments, where they are found to quantitatively rival commercial silicon diodes.

A review: Blood pressure monitoring based on PPG and circadian rhythm.

The demand for ambulatory blood pressure monitoring (ABPM) is increasing due to the global rise in cardiovascular disease patients. However, conventional ABPM methods are discontinuous and can disrupt daily activities and sleep patterns. Photoplethysmography (PPG) is gaining attention from researchers due to its simplicity, portability, affordability, and ease of signal acquisition. This paper critically examines the advancements achieved in the technology of PPG-guided noninvasive blood pressure (BP) monitoring and explores future opportunities. We have performed a literature search using the Web of Science and PubMed search engines, from January 2018 to October 2023, for PPG signal quality assessment (SQA), cuffless BP estimation using single PPG, and associations between circadian rhythm and BP. Based on this foundation, we first examine the impact of PPG signal quality on blood pressure estimation results while focusing on methods for assessing PPG signal quality. Subsequently, the methods documented for estimating cuff-free BP from PPG signals are summarized. Furthermore, the study examines how individual differences affect the accuracy of BP estimation, incorporating the factors that influence arterial blood pressure (ABP) and elucidating the impact of circadian rhythm on blood pressure. Finally, there will be a summary of the study's findings and suggestions for future research directions.

Design and validation of dual-point time-differentiated photoplethysmogram (2PPG) wearable for cuffless blood pressure estimation

Background & objectivesMeasurement of blood pressure (BP) in ambulatory patients is crucial for at high-risk cardiovascular patients. A non-obtrusive, non-occluding device that continuously measures BP via photoplethysmography will enable long-term ambulatory assessment of BP. The aim of this study is to validate the metasense 2PPG cuffless wearable design for continuous BP estimation without ECG. MethodsA customized high-speed electronic optical sensor architecture with laterally spaced reflectance pulse oximetry was designed into a simple unobtrusive low-power wearable in the form of a watch. 78 volunteers with a mean age of 32.72 ± 7.4 years (21 to 64), 51% male, 49% female were recruited with ECG-2PPG signals acquired. The fiducial features of the 2PPG morphologies were then attributed to the estimator. A 9–1 K-fold cross-validation was applied in the ML. ResultsThe correlation for PTT-SBP was 0.971 and for PTT-DBP was 0.954. The mean absolute error was 3.167±1.636 mmHg for SBP and 6.4 ± 3.9 mm Hg for DBP. The ambulatory estimate for SBP and DBP for an individual over 3 days with 8-hour recordings was 0.70–0.81 for SBP and 0.42–0.51 for DBP with a ± 2.65 mmHg for SBP and ±2.02 mmHg for DBP. For SBP, 98% of metasense measurements were within 15 mm Hg and for DBP, 91% of metasense measurements were within 10 mmHg ConclusionsThe metasense device provides continuous, non-invasive BP estimations that are comparable to ambulatory BP meters. The portability and unobtrusiveness of this device, as well as the ability to continuously measure BP could one day enable long-term ambulatory BP measurement for precision cardiovascular therapeutic regimens.

Advancing cuffless blood pressure estimation: A PPG-based multi-task learning model for enhanced feature extraction and fusion

Cuffless continuous blood pressure (BP) monitoring is essential for personalized health management. Although existing cuffless BP estimation applies advanced machine learning techniques and integrates PPG signals, it is deficient in feature extraction and fusion. In addition, it is inefficient to train the model separately for different tasks. In this study, an advanced multi-domain and local–global feature parallel multi-task learning network (MDLG-MTLNet) is introduced. The MDLG-MTLNet was designed with three key aspects: first, temporal and multi-scale spatial features were extracted from PPG signals and their derivatives via a multi-scale spatial and temporal feature block; interaction of features from different domains was facilitated by the introduction of a local–global attention module that captured and efficiently fused local–global information; and lastly, intrinsic correlation between systolic (SBP) and diastolic blood pressure (DBP) was taken into account via a multi-task learning network to improve the overall performance of the model. On the MIMIC-II dataset, the MAEs of MDLG-MTLNet for DBP, SBP, and MBP were 2.64 mmHg, 1.57 mmHg, and 2.02 mmHg, respectively. These errors were superior to those of the existing methods, meeting the AAMI criteria, and earning an A grade according to the BHS protocol. The experimental results confirm the effectiveness of our proposed model in feature extraction and fusion.

CiGNN: A Causality-Informed and Graph Neural Network Based Framework for Cuffless Continuous Blood Pressure Estimation.

Causalityholds profound potentials to dissipate confusion and improve accuracy in cuffless continuous blood pressure (BP) estimation, an area often neglected in current research. In this study, we propose a two-stage framework, CiGNN, that seamlessly integrates causality and graph neural network (GNN) for cuffless continuous BP estimation. The first stage concentrates on the generation of a causal graph between BP and wearable features from the the perspective of causal inference, so as to identify features that are causally related to BP variations. This stage is pivotal for the identification of novel causal features from the causal graph beyond pulse transit time (PTT). We found these causal features empower better tracking in BP changes compared to PTT. For the second stage, a spatio-temporal GNN (STGNN) is utilized to learn from the causal graph obtained from the first stage. The STGNN can exploit both the spatial information within the causal graph and temporal information from beat-by-beat cardiac signals for refined cuffless continuous BP estimation. We evaluated the proposed method with three datasets that include 305 subjects (102 hypertensive patients) with age ranging from 20-90 and BP at different levels, with the continuous Finapres BP as references. The mean absolute difference (MAD) for estimated systolic blood pressure (SBP) and diastolic blood pressure (DBP) were 3.77 mmHg and 2.52 mmHg, respectively, which outperformed comparison methods. In all cases including subjects with different age groups, while doing various maneuvers that induces BP changes at different levels and with or without hypertension, the proposed CiGNN method demonstrates superior performance for cuffless continuous BP estimation. These findings suggest that the proposed CiGNN is a promising approach in elucidating the causal mechanisms of cuffless BP estimation and can substantially enhance the precision of BP measurement.

Schrödinger spectrum and slim CNN architecture-based signal quality estimation for Photoplethysmogram signals

Photoplethysmography (PPG) signal comprises physiological information related to cardiorespiratory health. However, these signals are often contaminated by motion artifacts, resulting in low-quality noise-enriched signals. Therefore, it is crucial to ensure high-quality signals to extract accurate cardiorespiratory information. Although several rule-based and Machine-Learning (ML) - based approaches exist for PPG signal quality estimation but their effectiveness is questionable during evaluation in diverse datasets. This study proposes a lightweight, slim CNN architecture that utilizes a novel Quantum Pattern Recognition (QPR) technique for signal quality assessment. The proposed algorithm is initially validated on the University of Queensland database (clinical) and a noisy PPG database (non-clinical) collected from the ‘Welltory app’. A total of 28366, 5 s signal segments are preprocessed and transformed into image files of 20 × 500 pixels. The image files and heart rate obtained from the PPG signals are treated as input to the 2D slim CNN architecture. The developed model classifies the PPG signal as ‘good’ or ‘bad’ with an accuracy of 98.5%, 99.2% sensitivity, 97.5% specificity, and 99% F1-score. The classified database is further used for cuffless Blood Pressure estimation, and estimation error is reduced ≥ 1 mmHg w.r.t. reference BP signals. Finally, the proposed scheme is validated in STM32 series board for a resource-constrained wearable implementation. The computational load of the algorithm is reduced by network pruning and quantization for edge-device implementation, and it achieves an accuracy of 94.3%. This extensive analysis concludes that a slim architecture, along with a novel spatio–temporal pattern recognition technique, improves the system’s performance.

Validation of deep learning models for cuffless blood pressure estimation on a large benchmarking dataset

Objective: This study aims to evaluate the effectiveness of deep learning techniques in estimating cuffless blood pressure (BP) across a diverse patient population in intensive care units (ICUs). Methods: A comprehensive ICU benchmarking dataset encompassing 2,154 patients with a wide demographic range (18-97 years old) and varied cardiovascular status was employed to validate several deep learning models in predicting continuous BP waveforms. Three methods were developed to enhance the model's generalizability to this heterogeneous dataset. Ten-fold subject-independent cross-validation was performed and the model performance was assessed through mean absolute error (MAE), Pearson’s correlation coefficient (PCC), and compared with significance analysis. Results: The UTransBPNet_Demo_In model, which incorporated demographic and physiological signals as inputs, achieved a PCC of 0.89 and a MAE of 10.38 mmHg in predicting arterial BP waveforms, demonstrating the highest tracking capability among all models. Notably, the performance of UTransBPNet_Demo_In remained robust across variations in demographic and cardiovascular status. Conclusion: The UTransBPNet_Demo_In model demonstrates robust predictive capabilities across a broad spectrum of demographics and cardiovascular conditions. Although the performance still needs further improvement, this study offers a benchmark in the field of cuffless BP monitoring in critical care settings for future studies.

Cuffless blood pressure estimation from photoplethysmography using deep convolutional neural network and transfer learning

Blood pressure (BP) monitoring is an essential indicator for diseases of the cardiovascular system. Early detection of abnormalities in BP helps to significantly reduce the risk of diseases such as chronic heart failure, stroke and hypertension. In this study, we propose a new framework for systolic, diastolic and mean arterial blood pressure (SBP, DBP and MBP) estimation using PPG signals and its derivatives. The proposed framework includes the extraction of deep features by pre-trained CNN models based on transfer learning using images of signal waveforms. The most distinctive features are obtained by using the RFE feature selection algorithm on the extracted deep features. The selected deep features are exposed to a range of well-known machine and deep learning regression methods. The proposed BP estimation models have been evaluated with statistical metrics, visual analytical tools and international gold standards such as AAMI and BHS. Experimental results show that the first derivative of PPG, the VPG input image, the deep features obtained with DenseNet121 and the bi-directional gated recurrent unit (Bi-GRU) algorithm provide the best BP estimation performance. The results reveal that the SBP, DBP and MBP estimation models are grade-A according to the BHS protocol and meet the AAMI standard. The presented framework has been compared with well-established studies using the MIMIC II dataset, and the comparative analysis confirms that the proposed method outperforms existing techniques. The deep learning model for BP estimation offers a highly accurate, fast and practical system, especially in the monitoring of patients at risk of hypertension.

Variational Autoencoders for Biomedical Signal Morphology Clustering and Noise Detection.

Accurate estimation of physiological biomarkers using raw waveform data from non-invasive wearable devices requires extensive data preprocessing. An automatic noise detection method in time-series data would offer significant utility for various domains. As data labeling is onerous, having a minimally supervised abnormality detection method for input data, as well as an estimation of the severity of the signal corruptness, is essential. We propose a model-free, time-series biomedical waveform noise detection framework using a Variational Autoencoder coupled with Gaussian Mixture Models, which can detect a range of waveform abnormalities without annotation, providing a confidence metric for each segment. Our technique operates on biomedical signals that exhibit periodicity of heart activities. This framework can be applied to any machine learning or deep learning model as an initial signal validator component. Moreover, the confidence score generated by the proposed framework can be incorporated into different models' optimization to construct confidence-aware modeling. We conduct experiments using dynamic time warping (DTW) distance of segments to validated cardiac cycle morphology. The result confirms that our approach removes noisy cardiac cycles and the remaining signals, classified as clean, exhibit a 59.92% reduction in the standard deviation of DTW distances. Using a dataset of bio-impedance data of 97885 cardiac cycles, we further demonstrate a significant improvement in the downstream task of cuffless blood pressure estimation, with an average reduction of 2.67 mmHg root mean square error (RMSE) of Diastolic Blood pressure and 2.13 mmHg RMSE of systolic blood pressure, with increases of average Pearson correlation of 0.28 and 0.08, with a statistically significant improvement of signal-to-noise ratio respectively in the presence of different synthetic noise sources. This enables burden-free validation of wearable sensor data for downstream biomedical applications.

BrainZ-BP: A Noninvasive Cuff-Less Blood Pressure Estimation Approach Leveraging Brain Bio-Impedance and Electrocardiogram

BrainZ-BP: A Noninvasive Cuff-Less Blood Pressure Estimation Approach Leveraging Brain Bio-Impedance and Electrocardiogram

Photoplethysmography-based cuffless blood pressure estimation: an image encoding and fusion approach

Objective. Photoplethysmography (PPG) is a promising wearable technology that detects volumetric changes in microcirculation using a light source and a sensor on the skin’s surface. PPG has been shown to be useful for non-invasive blood pressure (BP) measurement. Deep learning-based BP measurements are now gaining popularity. However, almost all methods focus on 1D PPG. We aimed to design an end-to-end approach for estimating BP using image encodings from a 2D perspective. Approach. In this paper, we present a BP estimation approach based on an image encoding and fusion (BP-IEF) technique. We convert the PPG into five image encodings and use them as input. The proposed BP-IEF consists of two parts: an encoder and a decoder. In addition, three kinds of well-known neural networks are taken as the fundamental architecture of the encoder. The decoder is a hybrid architecture that consists of convolutional and fully connected layers, which are used to fuse features from the encoder. Main results. The performance of the proposed BP-IEF is evaluated on the UCI database in both non-mixed and mixed manners. On the non-mixed dataset, the root mean square error and mean absolute error for systolic BP (SBP) are 13.031 mmHg and 9.187 mmHg respectively, while for diastolic BP (DBP) they are 5.049 mmHg and 3.810 mmHg. On the mixed dataset, the corresponding values for SBP are 4.623 mmHg and 3.058 mmHg, while for DBP the values are 2.350 mmHg and 1.608 mmHg. In addition, both SBP and DBP estimation on the mixed dataset achieved grade A compared to the British Hypertension Society standard. The DBP estimation on the non-mixed dataset also achieved grade A. Significance. The results indicate that the proposed approach has the potential to improve on the current mobile healthcare for cuffless BP measurement.

Schrödinger spectrum based continuous cuff-less blood pressure estimation using clinically relevant features from PPG signal and its second derivative

The presented study estimates cuff-less blood pressure (BP) from photoplethysmography (PPG) signals using multiple machine-learning (ML) models and the semi-classical signal analysis (SCSA) technique. The study proposes a novel signal reconstruction algorithm, which optimizes the semi-classical constant of the SCSA approach and eliminates the trade-off between complexity and accuracy during signal reconstruction. It predicts BP values using regression algorithms by processing reconstructed PPG signals’ spectral features, extracting clinically relevant PPG and its second derivative’s (SDPPG) morphological features. The developed method was assessed using a virtual in-silico dataset with more than 4000 subjects and the Multi-Parameter Intelligent Monitoring in Intensive Care Units (MIMIC-II) dataset. Results showed that the method attained a mean absolute error (MAE) of 5.37 and 2.96 mmHg for systolic and diastolic BP, respectively, using the CatBoost algorithm. This approach met the Association for the Advancement of Medical Instrumentation’s standard and achieved Grade A for all BP categories in the British Hypertension Society protocol. The proposed framework performs well even when applied to a combined clinically relevant database originating from MIMIC-III and the Queensland dataset. The proposed method’s performance is further evaluated in a non-clinical setting with noisy and deformed PPG signals to validate the efficacy of the SCSA method. The noise stress tests further showed that the algorithm maintained its key feature detection, signal reconstruction capability, and estimation accuracy up to a 10 dB SNR ratio. The proposed cuff-less BP estimation technique has the potential to perform well in mobile healthcare devices due to its straightforward implementation approach.

Effects of Pulse Transit Time and Pulse Arrival Time on Cuff-less Blood Pressure Estimation: A Comparison Study with Multiple Experimental Interventions.

Pulse transit time (PTT) has shown a correlation with blood pressure (BP), and it is considered as a potential marker for cuff-less BP estimation. However, pulse arrival time (PAT) including pre-ejection period (PEP) has been utilized more widely because of its convenience to acquisition and calculation. In spite of this, whether PAT can surrogate PTT has been a controversial topic for many years. In this study, we designed an experiment on 55 subjects with multiple interventions, those may cause the changes in BP and PEP. We analyzed the linear and nonlinear correlations between BP and PTT/PAT, and also assessed the performances of PTT-based and PAT-based models on tracking the BP variation. Five typical BP estimation models were used for comparison. We found that PEP could change rapidly in response to the interventions related with physical stress. Although PTT had a better linear correlation with BP, most of the PAT-based models showed more accuracy than PTT-based models in all of the interventions, especially for the calibrated models. It is suggested that PAT has the potential to predict BP, and the inclusion of PEP in the measurement of PAT is necessary.

Continual Learning for Cuffless Blood Pressure Measurement using PPG and ECG Signals.

Although numerous studies have been conducted on cuffless blood pressure (BP) estimation using machine learning methods, most of the data-driven models are static, with model parameters fixed after training is complete. However, BP is dynamic and the performance would degrade for a static model when the to-be predicted BP distribution deviates from the training BP distribution. In this paper, we propose a continual learning (CL) framework in which deep learning models are developed to learn dynamically and continuously for arterial BP (ABP) estimation with photoplethysmography (PPG) and electrocardiogram (ECG) waveforms. The effectiveness of the CL model is validated on UCI Repository and MIMIC-III database with a total of 132 individual samples, and compared with conventional training method. It was found that the CL model improved the ABP estimation accuracy in terms of mean absolute error (MAE) by 17.47% on average compared with conventional training model. Furthermore, the improvement increased with the variability of ABP. These results demonstrate that CL model has potential to estimate dynamic ABP, which has been challenging with conventional training.

Boosting Algorithms based Cuff-less Blood Pressure Estimation from Clinically Relevant ECG and PPG Morphological Features.

Blood Pressure (BP) is often coined as a critical physiological marker for cardiovascular health. Multiple studies have explored either Photoplethysmogram (PPG) or ECG-PPG derived features for continuous BP estimation using machine learning (ML); deep learning (DL) techniques. Majority of those derived features often lack a stringent biological explanation and are not significantly correlated with BP. In this paper, we identified several clinically relevant (bio-inspired) ECG and PPG features; and exploited them to estimate Systolic (SBP), and Diastolic Blood Pressure (DBP) values using CatBoost, and AdaBoost algorithms. The estimation performance was then compared against popular ML algorithms. SBP and DBP achieved a Pearson's correlation coefficient of 0.90 and 0.83 between estimated and target BP values. The estimated mean absolute error (MAE) values are 3.81 and 2.22 mmHg with a Standard Deviation of 6.24 and 3.51 mmHg, respectively, for SBP and DBP using CatBoost. The results surpassed the Advancement of Medical Instrumentation (AAMI) standards. For the British Hypertension Society (BHS) protocol, the results achieved for all the BP categories resided in Grade A. Further investigation reveals that bio-inspired features along with tuned ML models can produce comparable results w.r.t parameter-intensive DL networks. ln(HR × mNPV), HR, BMI index, ageing index, and PPG-K point were identified as the top five key features for estimating BP. The group-based analysis further concludes that a trade-off lies between the number of features and MAE. Increasing the no. of features beyond a certain threshold saturates the reduction in MAE.

Investigation of Data Leakage in Deep-Learning-Based Blood Pressure Estimation Using Photoplethysmogram/Electrocardiogram

Cuff-less blood pressure (BP) estimation methods using a deep learning model from photoplethysmogram (PPG) or electrocardiogram (ECG) have been actively studied in recent years. However, we found that most previous studies incur data leakage, where segments or records measured from the same subject appear in both the training and test datasets. Furthermore, many previous studies are suspected to have misinterpreted a record in the public dataset used for their evaluations as a subject. To investigate data leakage in BP estimation methods, this paper first organizes previous studies in terms of data leakage. We then quantitatively evaluate the effect of data leakage caused by the segment-level and the record-level train-test split using the public dataset, Cuff-Less Blood Pressure Estimation Data Set. Our experimental results showed that the segment-level split and record-level split erroneously improved the estimation accuracy of mean blood pressure from the (quasi-)subject-level split by the Pearson’s correlation coefficient of 0.56 and 0.40 when using PPG, and 0.82 and 0.69 when using ECG, respectively. These results confirmed that the train-test split used in many previous studies, including the one that described its evaluation as causing no data leakage, causes a high level of data leakage, and that a record in the Cuff-Less Blood Pressure Estimation Data Set, often misinterpreted as a subject in previous studies, is not a subject and that a high level of data leakage occurs when the record is considered as a subject.