GRU Neural Network Improved Bioimpedance Based Stroke Volume Estimation during Ergometry Stress Test.

Cardiovascular diseases (CVDs) are one of the leading members of non-communicable diseases. An early diagnosis is essential for effective treatment, to reduce hospitalization time and health care costs. Nowadays, an exercise stress test on an ergometer is used to identify CVDs. To improve the accuracy of diagnostics, the hemodynamic status and parameters of a person can be investigated. For hemodynamic management, thoracic electrical bioimpedance has recently been used. This technique offers beat-to-beat stroke volume calculation but suffers from an artifact-sensitive signal that makes such measurements difficult during movement. We propose a new method based on a gated recurrent unit (GRU) neural network and the ECG signal to improve the measurement of bioimpedance signals, reduce artifacts and calculate hemodynamic parameters. We conducted a study with 23 subjects. The new approach is compared to ensemble averaging, scaled Fourier linear combiner, adaptive filter, and simple neural networks. The GRU neural network performs better with single artifact events than shallow neural networks (mean error -0.0244, mean square error 0.0181 for normalized stroke volume). The GRU network is superior to other algorithms using time-correlated data for the exercise stress test.

Detection of a Stroke Volume Decrease by Machine-Learning Algorithms Based on Thoracic Bioimpedance in Experimental Hypovolaemia.

Compensated shock and hypovolaemia are frequent conditions that remain clinically undetected and can quickly cause deterioration of perioperative and critically ill patients. Automated, accurate and non-invasive detection methods are needed to avoid such critical situations. In this experimental study, we aimed to create a prediction model for stroke volume index (SVI) decrease based on electrical cardiometry (EC) measurements. Transthoracic echo served as reference for SVI assessment (SVI-TTE). In 30 healthy male volunteers, central hypovolaemia was simulated using a lower body negative pressure (LBNP) chamber. A machine-learning algorithm based on variables of EC was designed. During LBNP, SVI-TTE declined consecutively, whereas the vital signs (arterial pressures and heart rate) remained within normal ranges. Compared to heart rate (AUC: 0.83 (95% CI: 0.73-0.87)) and systolic arterial pressure (AUC: 0.82 (95% CI: 0.74-0.85)), a model integrating EC variables (AUC: 0.91 (0.83-0.94)) showed a superior ability to predict a decrease in SVI-TTE ≥ 20% (p = 0.013 compared to heart rate, and p = 0.002 compared to systolic blood pressure). Simulated central hypovolaemia was related to a substantial decline in SVI-TTE but only minor changes in vital signs. A model of EC variables based on machine-learning algorithms showed high predictive power to detect a relevant decrease in SVI and may provide an automated, non-invasive method to indicate hypovolaemia and compensated shock.

Surrogate based continuous noninvasive blood pressure measurement.

Arterial blood pressure is one of the most often measured vital parameters in clinical practice. State-of-the-art noninvasive ABP measurement technologies have noticeable limitations and are mainly based on uncomfortable techniques of complete or partial arterial occlusion by cuffs. Most commonplace devices provide only intermittent measurements, and continuous systems are bulky and difficult to apply correctly for nonprofessionals. Continuous cuffless ABP measurements are still an unmet clinical need and a topic of ongoing research, with only few commercially available devices. This paper discusses surrogate-based noninvasive blood pressure measurement techniques. It covers measurement methods of continuously and noninvasively inferring BP from surrogate signals without applying external pressures, except for reference or initialization purposes. The BP is estimated by processing signal features, so called surrogates, which are modulated by variations of BP. Discussed techniques include well-known approaches such as pulse transit time and pulse arrival time techniques, pulse wave analysis or combinations thereof. Despite a long research history, these methods have not found widespread use in clinical and ambulatory practice, in part due to technical limitations and the lack of a standardized regulatory framework. This work summarizes findings from an invited workshop of experts in the fields covering clinical expertise, engineering aspects, commercialization and standardization issues. The goal is to provide an application driven outlook, starting with clinical needs, and extending to technical actuality. It provides an outline of recommended research directions and includes a detailed overview of clinical use case scenarios for these technologies, opportunities, and limitations.

Wearable Cardiorespiratory Monitoring Employing a Multimodal Digital Patch Stethoscope: Estimation of ECG, PEP, LVETand Respiration Using a 55 mm Single-Lead ECG and Phonocardiogram.

Cardiovascular diseases are the main cause of death worldwide, with sleep disordered breathing being a further aggravating factor. Respiratory illnesses are the third leading cause of death amongst the noncommunicable diseases. The current COVID-19 pandemic, however, also highlights the impact of communicable respiratory syndromes. In the clinical routine, prolonged postanesthetic respiratory instability worsens the patient outcome. Even though early and continuous, long-term cardiorespiratory monitoring has been proposed or even proven to be beneficial in several situations, implementations thereof are sparse. We employed our recently presented, multimodal patch stethoscope to estimate Einthoven electrocardiogram (ECG) Lead I and II from a single 55 mm ECG lead. Using the stethoscope and ECG subsystems, the pre-ejection period (PEP) and left ventricular ejection time (LVET) were estimated. ECG-derived respiration techniques were used in conjunction with a novel, phonocardiogram-derived respiration approach to extract respiratory parameters. Medical-grade references were the SOMNOmedics SOMNO HDTM and Osypka ICON-CoreTM. In a study including 10 healthy subjects, we analyzed the performances in the supine, lateral, and prone position. Einthoven I and II estimations yielded correlations exceeding 0.97. LVET and PEP estimation errors were 10% and 21%, respectively. Respiratory rates were estimated with mean absolute errors below 1.2 bpm, and the respiratory signal yielded a correlation of 0.66. We conclude that the estimation of ECG, PEP, LVET, and respiratory parameters is feasible using a wearable, multimodal acquisition device and encourage further research in multimodal signal fusion for respiratory signal estimation.

Balanced Adjustable Mirrored Current Source with Common Mode Feedback and Output Measurement for Bioimpedance Applications.

Bioimpedance methods are used in a variety of applications such as impedance tomography, electrodermal activity detection and vascular disease assessment. Recent developments in portable and unobtrusive biosignal acquisition systems facilitate the integration of wearable bioimpedance applications including sleep monitoring, respiration estimation and fluid monitoring. However, the less stable measurement situation in a wearable scenario increases the requirements for the system's accuracy and adaptability. The current source of a bioimpedance system needs to drive large complex loads subject to vast variations over time while maintaining a high level of accuracy. The widely used improved Howland current source suffers from multiple disadvantages when considered for an adaptive bioimpedance system. We propose an optimized mirrored architecture which allows for a simple output current adjustment and current measurement without an additional shunt resistor in the load path. The system implements a common mode feedback system which includes balancing of the mirrored sources. Our design is validated by calculation, SPICE simulation and complex load measurements. We achieved output impedances in excess of 3 MΩ and derived a simplified transconductance function valid for frequencies up to 1 MHz. We conclude that the presented architecture is an important step forward towards accurate wearable bioimpedance acquisition. Employing generalized impedance converters, the output impedance could be further optimized.

Evaluation of Pulse Arrival Times during Lower Body Negative Pressure Test for the Non-Invasive Detection of Hypovolemia.

The early detection of occult bleeding is a difficult problem for clinicians because physiological variables such as heart rate and blood pressure that are measured with standard patient monitoring equipment are insensitive to blood loss. In this study, the pulse arrival time (PAT) was investigated as an easily recorded, non-invasive indicator of hypovolemia. A lower body negative pressure (LBNP) study with a stepwise increase of negative pressure was conducted to induce central hypovolemia in a study population of 30 subjects. PAT values were extracted from simultaneous recordings of the electrocardiogram (ECG) and photoplethysmographic (PPG) recordings both from the index finger and from within the outer ear canal. Stroke volume (SV) was recorded as a reference measure by transthoracic echocardiography. An inter- and intra-individual correlation analysis between changes in SV and the PAT measurements was performed. Furthermore, it was assessed if PAT measurements can indicate a diminished SV in this scenario. It could be demonstrated that the measured PAT values are significantly increased at the lowest LBNP pressure level. A very strong intra-individual correlation (ρ ≥ 0.8) and a moderate inter-individual correlation (ρ ≥ 0.5) between PAT and SV measurements were found. Thus, PAT measurements could be a viable tool to monitor patient specific volemic trends. Further research is needed to investigate if PAT information can be utilized for a more robust inter-subject quantification of the degree of hypovolemia.

Wearable Multimodal Stethoscope Patch for Wireless Biosignal Acquisition and Long-Term Auscultation.

Detecting critical events in postoperative care and improving comfort, costs and availability in sleep assessment are two of many areas in which wearable biosignal acquisition can be a viable tool. Modern sensors as well as patch and textile integration facilitate unobtrusive biosignal acquisition, yet placing sensors at different locations across the body is still prevailing. Actigraphy and the electrocardiogram (ECG) are commonly integrated modalities. The stethoscope however, despite its wide range of applications, has been neglected from these developments. The introduction of digital stethoscopes, recently led to an objectification and increased interest in the field. We present the prototype of a wearable, Bluetooth 5.0 LE enabled multimodal sensor patch combining five modalities: MEMS stethoscope, ambient noise sensing, ECG, impedance pneumography (IP) and 9-axial actigraphy. The system alleviates the need for sensors at different body positions and enables long-term auscultation. Using high sampling rates and online synchronization, multimodal sensor fusion becomes feasible. The patch measures 70 mm x 60 mm and is attached using three 24 mm Ag/AgCl electrodes. High quality cardiac and pulmonary auscultation as well as ECG and IP acquisition are demonstrated. We derived respiration surrogates with linear correlations to a reference exceeding 0.91 and conclude that the system can be utilized in fields requiring unobtrusive yet high quality signal acquisition. Future research will include the integration of additional sensors and further size reduction.

Model selection for the Pulse Decomposition Analysis of fingertip photoplethysmograms.

In the analysis of fingertip photoplethysmograms (PPG), the Pulse Decomposition Analysis (PDA) has emerged as a powerful tool for the extraction of physiologically relevant information from the morphology of single digital volume pulse (DVP) cycles. In previously published works on the PDA, many different models are suggested. In this work, we conducted a data driven approach to address the question of which model to choose for the PDA. For this purpose, we compiled an extensive dataset of 7805 single DVP pulses that comprises most expectable pulse morphologies and conducted PDA simulations with four different basis functions types and a meaningful range of model orders. We then performed model selection based on the Corrected Akaike Information Criterion (AICc) with the aim of identifying the PDA models that provided the best fit. As a result, we found that a PDA model based on the linear superposition of three scaled Gamma basis functions was selected as the best fitting model in 28.1% of all pulses. The second highest relative selection frequency of 14.4% was achieved by fitting two Rayleigh functions. Consequently, we recommend to consider the employment of this PDA model in further work on the PDA.

A Fast Multimodal Ectopic Beat Detection Method Applied for Blood Pressure Estimation Based on Pulse Wave Velocity Measurements in Wearable Sensors.

Automatic detection of ectopic beats has become a thoroughly researched topic, with literature providing manifold proposals typically incorporating morphological analysis of the electrocardiogram (ECG). Although being well understood, its utilization is often neglected, especially in practical monitoring situations like online evaluation of signals acquired in wearable sensors. Continuous blood pressure estimation based on pulse wave velocity considerations is a prominent example, which depends on careful fiducial point extraction and is therefore seriously affected during periods of increased occurring extrasystoles. In the scope of this work, a novel ectopic beat discriminator with low computational complexity has been developed, which takes advantage of multimodal features derived from ECG and pulse wave relating measurements, thereby providing additional information on the underlying cardiac activity. Moreover, the blood pressure estimations' vulnerability towards ectopic beats is closely examined on records drawn from the Physionet database as well as signals recorded in a small field study conducted in a geriatric facility for the elderly. It turns out that a reliable extrasystole identification is essential to unsupervised blood pressure estimation, having a significant impact on the overall accuracy. The proposed method further convinces by its applicability to battery driven hardware systems with limited processing power and is a favorable choice when access to multimodal signal features is given anyway.

Indirect Fourier transform in the context of statistical inference.

Inferring structural information from the intensity of a small-angle scattering (SAS) experiment is an ill-posed inverse problem. Thus, the determination of a solution is in general non-trivial. In this work, the indirect Fourier transform (IFT), which determines the pair distance distribution function from the intensity and hence yields structural information, is discussed within two different statistical inference approaches, namely a frequentist one and a Bayesian one, in order to determine a solution objectively From the frequentist approach the cross-validation method is obtained as a good practical objective function for selecting an IFT solution. Moreover, modern machine learning methods are employed to suppress oscillatory behaviour of the solution, hence extracting only meaningful features of the solution. By comparing the results yielded by the different methods presented here, the reliability of the outcome can be improved and thus the approach should enable more reliable information to be deduced from SAS experiments.

Multi-modal signal acquisition using a synchronized wireless body sensor network in geriatric patients.

Wearable home-monitoring devices acquiring various biosignals such as the electrocardiogram, photoplethysmogram, electromyogram, respirational activity and movements have become popular in many fields of research, medical diagnostics and commercial applications. Especially ambulatory settings introduce still unsolved challenges to the development of sensor hardware and smart signal processing approaches. This work gives a detailed insight into a novel wireless body sensor network and addresses critical aspects such as signal quality, synchronicity among multiple devices as well as the system's overall capabilities and limitations in cardiovascular monitoring. An early sign of typical cardiovascular diseases is often shown by disturbed autonomic regulations such as orthostatic intolerance. In that context, blood pressure measurements play an important role to observe abnormalities like hypo- or hypertensions. Non-invasive and unobtrusive blood pressure monitoring still poses a significant challenge, promoting alternative approaches including pulse wave velocity considerations. In the scope of this work, the presented hardware is applied to demonstrate the continuous extraction of multi modal parameters like pulse arrival time within a preliminary clinical study. A Schellong test to diagnose orthostatic hypotension which is typically based on blood pressure cuff measurements has been conducted, serving as an application that might significantly benefit from novel multi-modal measurement principles. It is further shown that the system's synchronicity is as precise as 30 µs and that the integrated analog preprocessing circuits and additional accelerometer data provide significant advantages in ambulatory measurement environments.

Improved signal quality indication for photoplethysmographic signals incorporating motion artifact detection.

Wearable monitoring systems have gained tremendous popularity in the health-care industry, opening new possibilities in diagnostic routines and medical treatments. Numerous hardware systems have been presented since, which allow for continuous acquisition of various biosignals like the ECG, PPG, EMG or EEG and which are suited for ambulatory settings. Unfortunately, these flexible systems are liable to motion artifacts and especially photoplethysmographic signals are seriously distorted when the patient is not at rest. A lot of work has been done to reduce artifacts and noise, ranging from simple filtering methods to very complex statistical approaches. With regard to the PPG, certain quality indices have been proposed to evaluate the signal conditions. As movements are the primary source of signal disturbances, the relation between the output of a signal quality estimator and acceleration data captured directly on the PPG sensor is focused in this work. It will be shown that typical motions can be detected on-line, thereby providing additional information which will significantly improve signal quality assessments.

Online learning algorithms for principal component analysis applied on single-lead ECGs.

This article evaluates several adaptive approaches to solve the principal component analysis (PCA) problem applied on single-lead ECGs. Recent studies have shown that the principal components can indicate morphologically or environmentally induced changes in the ECG signal and can be used to extract other vital information such as respiratory activity. Special interest is focused on the convergence behavior of the selected gradient algorithms, which is a major criterion for the usability of the gained results. As the right choice of learning rates is very data dependant and subject to movement artifacts, a new measurement system was designed, which uses acceleration data to improve the performance of the online algorithms. As the results of PCA seem very promising, we propose to apply a single-channel independent component analysis (SCICA) as a second step, which is investigated in this paper as well.

Wireless Body Sensor Network for low-power motion-tolerant synchronized vital sign measurement.

Prophylaxis and rehabilitation of cardiovascular disease require the development of biosignal acquisition and processing devices that are capable of supporting patients in their everyday life. This paper presents a Body Sensor Network (BSN) for use in Personal Healthcare applications. It consists of miniaturized sensor modules for electrocardiogram (ECG), photoplethysmogram (PPG) and phonocardiography (PCG) which are wirelessly connected with a coordinator to collect the data. Each sensor module is combined with a tri-axis accelerometer for patient's posture and activity measurement. As it is possible to extract further information about the health state by fusioning data of different biosensors, the wireless link based on IEEE 802.15.4 was extended by a synchronisation mechanism enabling synchronous sampling of the individual sensors. An adaptive application of algorithms for signal pre-processing and analysis allows the reduction of the transferred data.