Respiratory Activity during Exercise: A Feasibility Study on Transition Point Estimation Using Impedance Pneumography

The current diagnostic procedures for assessing physiological response to exercise comprise blood lactates measurements, ergospirometry, and electrocardiography. The first is not continuous, the second requires specialized equipment distorting natural breathing, and the last is indirect. Therefore, we decided to perform the feasibility study with impedance pneumography as an alternative technique. We attempted to determine points in respiratory-related signals, acquired during stress test conditions, that suggest a transition similar to the gas exchange threshold. In addition, we analyzed whether or not respiratory activity reaches steady states during graded exercise. Forty-four students (35 females), practicing sports on different levels, performed a graded exercise test until exhaustion on cycloergometer. Eventually, the results from 34 of them were used. The data were acquired with Pneumonitor 2. The signals demonstrated that the steady state phenomenon is not as evident as for heart rate. The results indicated respiratory rate approaches show the transition point at the earliest (more than 6 min before the end of the exercise test on average), and the tidal volume ones at the latest (less than 5 min). A combination gave intermediate findings. The results showed the impedance pneumography appears reasonable for the transition point estimation, but this should be further studied with the reference.

Artefact Detection in Impedance Pneumography Signals: A Machine Learning Approach

Impedance pneumography has been suggested as an ambulatory technique for the monitoring of respiratory diseases. However, its ambulatory nature makes the recordings more prone to noise sources. It is important that such noisy segments are identified and removed, since they could have a huge impact on the performance of data-driven decision support tools. In this study, we investigated the added value of machine learning algorithms to separate clean from noisy bio-impedance signals. We compared three approaches: a heuristic algorithm, a feature-based classification model (SVM) and a convolutional neural network (CNN). The dataset consists of 47 chronic obstructive pulmonary disease patients who performed an inspiratory threshold loading protocol. During this protocol, their respiration was recorded with a bio-impedance device and a spirometer, which served as a gold standard. Four annotators scored the signals for the presence of artefacts, based on the reference signal. We have shown that the accuracy of both machine learning approaches (SVM: 87.77 ± 2.64% and CNN: 87.20 ± 2.78%) is significantly higher, compared to the heuristic approach (84.69 ± 2.32%). Moreover, no significant differences could be observed between the two machine learning approaches. The feature-based and neural network model obtained a respective AUC of 92.77±2.95% and 92.51±1.74%. These findings show that a data-driven approach could be beneficial for the task of artefact detection in respiratory thoracic bio-impedance signals.

Feasibility of non-contact cardiorespiratory monitoring using impulse-radio ultra-wideband radar in the neonatal intensive care unit

Background Current cardiorespiratory monitoring equipment can cause injuries and infections in neonates with fragile skin. Impulse-radio ultra-wideband (IR-UWB) radar was recently demonstrated to be an effective contactless vital sign monitor in adults. The purpose of this study was to assess heart rates (HRs) and respiratory rates (RRs) in the neonatal intensive care unit (NICU) using IR-UWB radar and to evaluate its accuracy and reliability compared to conventional electrocardiography (ECG)/impedance pneumography (IPG). Methods The HR and RR were recorded in 34 neonates between 3 and 72 days of age during minimal movement (51 measurements in total) using IR-UWB radar (HRRd, RRRd) and ECG/IPG (HRECG, RRIPG) simultaneously. The radar signals were processed in real time using algorithms for neonates. Radar and ECG/IPG measurements were compared using concordance correlation coefficients (CCCs) and Bland-Altman plots. Results From the 34 neonates, 12,530 HR samples and 3,504 RR samples were measured. Both the HR and RR measured using the two methods were highly concordant when the neonates had minimal movements (CCC = 0.95 between the RRRd and RRIPG, CCC = 0.97 between the HRRd and HRECG). In the Bland-Altman plot, the mean biases were 0.17 breaths/min (95% limit of agreement [LOA] -7.0–7.3) between the RRRd and RRIPG and -0.23 bpm (95% LOA -5.3–4.8) between the HRRd and HRECG. Moreover, the agreement for the HR and RR measurements between the two modalities was consistently high regardless of neonate weight. Conclusions A cardiorespiratory monitor using IR-UWB radar may provide accurate non-contact HR and RR estimates without wires and electrodes for neonates in the NICU.

Non-Contact Video-Based Neonatal Respiratory Monitoring

Respiratory rate (RR) has been shown to be a reliable predictor of cardio-pulmonary deterioration, but standard RR monitoring methods in the neonatal intensive care units (NICU) with contact leads have been related to iatrogenic complications. Video-based monitoring is a potential non-contact system that could improve patient care. This iterative design study developed a novel algorithm that produced RR from footage analyzed from stable NICU patients in open cribs with corrected gestational ages ranging from 33 to 40 weeks. The final algorithm used a proprietary technique of micromotion and stationarity detection (MSD) to model background noise to be able to amplify and record respiratory motions. We found significant correlation—r equals 0.948 (p value of 0.001)—between MSD and the current hospital standard, electrocardiogram impedance pneumography. Our video-based system showed a bias of negative 1.3 breaths and root mean square error of 6.36 breaths per minute compared to standard continuous monitoring. Further work is needed to evaluate the ability of video-based monitors to observe clinical changes in a larger population of patients over extended periods of time.

Simple Wireless Impedance Pneumography System for Unobtrusive Sensing of Respiration

This extended paper presents the development and implementation at a prototype level of a wireless, low-cost system for the measurement of the electrical bioimpedance of the chest with two channels using the AD5933 in a bipolar electrode configuration to measure impedance pneumography. The measurement device works for impedance measurements ranging from 1 Ω to 1800 Ω. Fifteen volunteers were measured with the prototype. We found that the left hemithorax has higher impedance compared to the right hemithorax, and the acquired signal presents the phases of the respiratory cycle with variations between 1 Ω, in normal breathing, to 6 Ω in maximum inhalation events. The system can measure the respiratory cycle variations simultaneously in both hemithorax with a mean error of −0.18 ± 1.42 BPM (breaths per minute) in the right hemithorax and −0.52 ± 1.31 BPM for the left hemithorax, constituting a useful device for the breathing rate calculation and possible screening applications.

Wearable Impedance Pneumography

Respiratory diseases are a leading cause of death worldwide. The prevalence of sleep apnea, its cardiovascular consequences, postoperative respiratory instability and severe respiratory syndromes further highlight the importance of respiratory monitoring. Typical methods, however, rely on obtrusive nasal cannulas and belts. Impedance pneumography (IP) is a promising bioimpedance application which aims to estimate respiratory parameters from the thorax impedance. Currently, IP configurations require large inter-electrode distances, diminishing its applicability in a wearable context. We propose an IP configuration with 55 mm spacing using our recently presented sensor patch. In a study including 10 healthy subjects, respiratory rate (RR) and flow are estimated in the supine, lateral and prone position. Using time-delay neural network regression, RR errors below 1 bpm, flow correlations of 0.81 and relative flow errors of 38 % with respect to a pneumotachometer reference were achieved. We conclude that high accuracy RR estimation is possible in a 55 mm IP configuration. Respiratory flow can be roughly estimated. Further research combining several biosignals for a more robust, wearable flow estimation is recommended.