key: cord-0752406-c1n1j8wm authors: Al-Halhouli, Ala’aldeen; Al-Ghussain, Loiy; Khallouf, Osama; Rabadi, Alexander; Alawadi, Jafar; Liu, Haipeng; Al Oweidat, Khaled; Chen, Fei; Zheng, Dingchang title: Clinical Evaluation of Respiratory Rate Measurements on COPD (Male) Patients Using Wearable Inkjet-Printed Sensor date: 2021-01-11 journal: Sensors (Basel) DOI: 10.3390/s21020468 sha: 7a9f71387799142a01d7d4f9fc9e3e70261e0958 doc_id: 752406 cord_uid: c1n1j8wm Introduction: Chronic Obstructive Pulmonary Disease (COPD) is a progressive disease that causes long-term breathing problems. The reliable monitoring of respiratory rate (RR) is very important for the treatment and management of COPD. Based on inkjet printing technology, we have developed a stretchable and wearable sensor that can accurately measure RR on normal subjects. Currently, there is a lack of comprehensive evaluation of stretchable sensors in the monitoring of RR on COPD patients. We aimed to investigate the measurement accuracy of our sensor on COPD patients. Methodology: Thirty-five patients (Mean ± SD of age: 55.25 ± 13.76 years) in different stages of COPD were recruited. The measurement accuracy of our inkjet-printed (IJPT) sensor was evaluated at different body postures (i.e., standing, sitting at 90°, and lying at 45°) on COPD patients. The RR recorded by the IJPT sensor was compared with that recorded by the reference e-Health sensor using paired T-test and Wilcoxon signed-rank test. Analysis of variation (ANOVA) was performed to investigate if there was any significant effect of individual difference or posture on the measurement error. Statistical significance was defined as p-value less than 0.05. Results: There was no significant difference between the RR measurements collected by the IJPT sensor and the e-Health reference sensor overall and in three postures (p > 0.05 in paired T-tests and Wilcoxon signed-rank tests). The sitting posture had the least measurement error of −0.0542 ± 1.451 bpm. There was no significant effect of posture or individual difference on the measurement error or relative measurement error (p > 0.05 in ANOVA). Conclusion: The IJPT sensor can accurately measure the RR of COPD patients at different body postures, which provides the possibility for reliable monitoring of RR on COPD patients. Respiratory rate (RR) is a vital sign that is related to, therefore regulated by, multiple physiological and neural activities [1] [2] [3] . RR plays an important role in the detection of various cardiovascular and respiratory diseases, as well as relevant clinical events [4, 5] . The variation of RR reflects the deterioration of respiratory diseases including asthma, chronic obstructive pulmonary disease (COPD), and other clinical conditions including The extraction of the RR using the IJPT on COPD patients consists of several stages, including the fabrication of the IJPT sensor and the signal processing of the RR signal. Figure 1 shows a brief summary of the fabrication process of the IJPT sensor as well as the RR extraction procedure on COPD patients. The following sections elaborate more on these stages. The protocol adhered to the tenets of the Declaration of Helsinki [43] . Each patient had the freedom of choice to participate in this work. The procedure was explained in details with questions answered. Then, the participants signed the consent form, which was reviewed and approved by Jordan University Hospital Ethics Committee (REF 67/2019/6480). The fabrication of flexible electronics using inkjet printing technology has gained significant interest in recent years. The RR sensor evaluated in this study was fabricated using this technology where silver nanoparticle ink was deposited on polydimethylsiloxane (PDMS) substrates. The RR was detected by the IJPT sensor via the variations of the resistance of the conductive traces caused by the volumetric change in the ribcage or abdomen areas during the respiration process. In other words, the IJPT sensor acted as a strain gauge sensor. The details of the fabrication process of the IJPT sensor can be found in [16] . Thirty-five patients were included in the study after taking their permissions using the informed consent form. All of them were diagnosed with COPD, and managed with adequate medication and on regular follow-up. Firstly, patients were asked to rest for five minutes while the questionnaire was filled with the health records in patient's file validated. Then, the respiration rate was measured with the reference e-Health nasal flow sensor and the IJPT sensor simultaneously. The measurement was repeated in 3 different positions: Lying position 45 • , sitting upright, and standing. In each position, the measurement lasted for one minute. The e-Health sensor was placed in the nostrils while the IJPT sensor was fixed on the xiphoid process through an adjustable belt over the clothes. It should be noted that the patients were asked not to move while taking the measurements. The patients included in the study had different GOLD stages of COPD (1, 2, 3 or 4), different age groups (mean ± standard deviation (SD) of age: 55.25 ± 13.76 years) and different smoking status (ex-smoker, quit smoking, or smoker). Some of the limitations faced prevented the measurement of the RR at some postures, such as the inability to take readings from patients on continuous oxygen support as the e-Health sensor was placed by the nostrils. Another limitation was the endurance of some patients to stay in a certain position as it caused fatigue and irritability. In addition, some patients had chest tubes, which caused improper mounting of the IJPT sensor. Despite these limitations, the respiratory signals were successfully recorded in most measurements. A summary of patients' information, as well as CAGE (Cut-Annoyed-Guilty-Eye) questionnaire summary, can be found in Tables A1 and A2 in Appendix A, respectively. The strain gauge sensor, whose resistance value varied during respiration, was connected to a Wheatstone Bridge with suitable values for the resistors. The output was connected to an Instrumentational Amplifier with a suitable value for the gain resistor, as shown in Figure 2 . The output of the amplifier was connected to an analog input of the Arduino board. The airflow sensor was connected to e-Health Sensor Shield V2.0 for Arduino [44] and connected to the same Arduino board as the strain gauge circuitry. Arduino Mega2560 with an ATMega2560 chip as its microcontroller was used with maximum sampling rate of 15 kSPS at maximum resolution [45, 46] . The signals of the strain gauge sensor and the airflow sensor were sampled at 100 Hz using the Arduino board, which was connected to a MATLAB/Simulink file to save the data. Afterward, a MATLAB-based bandpass filter was used to filter both signals with cut-off frequencies of 3 and 90 bpm, or 0.05 and 1.5 Hz, respectively, with the function configured to an infinite impulse response. The calculation of the respiratory rate was done by finding the frequency at which the power spectral density of the filtered data had its maximum value. Figure 3 shows the raw RR signals from the IJPT and e-Health sensors as well as power spectral densities of both signals before and after the filtering. The circle and cross indicate the maximum value of the power spectral density of each signal at which the respiratory rate was found. For the data of each measurement, the quality of reference (airflow) respiratory signal was double-checked. If the quality of reference respiratory signal was low (e.g., blurred waveform, missing period longer than 5 s, or missing more than two consecutive respiratory cycles), the data of that measurement were discarded. The considered RR data can be found in Table A3 in Appendix A. The RR values derived from inkjet-printed strain gauge (RR SG ) and airflow (RR AF ) were compared to investigate if there was any significant difference. Firstly, the Shapiro-Wilk test was performed on RR values to investigate if they followed normal distribution. If both RR SG and RR AF groups followed normal distribution (defined as p > 0.05 in Shapiro-Wilk test), the paired T-test was performed to investigate if there was any significant difference (defined as p < 0.05) between RR SG and RR AF . If normal distribution was not followed in any one group, the Wilcoxon signed-rank test (significant difference defined as p < 0.05) was performed as the non-parametric substitute of paired T-test. Considering the missing data in some subjects, the test was performed globally and for each posture, respectively. The estimation error of RR was calculated as in Equation. (1). The analysis of variance (ANOVA) was performed to investigate if there was any significant effect of individual difference or posture on the results. It has been proven that ANOVA is applicable even if the data deviate from normal distribution [47] . For reliable estimation, the Shapiro-Wilk test was performed to investigate if the error followed normal distribution in three postures respectively. The paired T-test was performed on the data of error derived in different postures if both groups followed normal distribution. Otherwise, the Kruskal-Wallis test was used to investigate if the effect of posture on error was significant (defined as p < 0.05), while the Wilcoxon signed-rank test was performed as the non-parametric substitute of paired T-test. The relative error of RR estimation was calculated as in Equation (2). The analysis method is the same as that for error of RR. To illustrate the difference between RR SG and RR AF in different postures, the Bland-Altman analysis was performed on RR SG and RR AF for each posture, respectively [48, 49] . In signal processing, the resolution of RR was 0.001 bpm. To accurately show and compare the biases, the results of Bland-Altman analysis were rounded to the third significant digit after the decimal point. Linear regression analysis was used to inspect whether the correlation between the RR SG and the RR AF followed a linear correlation or not. The consistency in the RR SG and RR AF was evaluated using the regression coefficient of the linear correlation. The overall distribution of RR SG and RR AF followed normal distribution (p = 0.689 and p = 0.066 in Shapiro-Wilk test). The paired T-test showed no significant difference between RR SG and RR AF (p = 0.572). In sitting posture, normal distribution was followed by RR AF but not RR SG (p = 0.104 and p = 0.045 in Shapiro-Wilk test, respectively). The Wilcoxon signed-rank test showed no significant difference between RR SG and RR AF (p = 0.968). In standing posture, both RR SG and RR AF followed normal distribution (p = 0.977 and p = 0.652 in Shapiro-Wilk test, respectively). The paired T-test showed no significant difference between RR SG and RR AF (p = 0.873). In lying45 • posture, both RR SG and RR AF (p = 0.629 and p = 0.242 in Shapiro-Wilk test, respectively) followed normal distribution. The paired T-test showed no significant difference between RR SG and RR AF (p = 0.619). The ANOVA showed no significant effect of posture (p = 0.318) or individual difference (p = 0.857) on the measurement error. The normal distribution was followed by the data of standing posture (p = 0.516 in Shapiro-Wilk test) but not those of sitting (p = 0.007) or lying45 • posture (p < 0.001). The Wilcoxon signed-rank test showed no significant difference between the data of sitting and standing (p = 0.754), sitting and lying45 • (p = 0.107), or standing and lying45 • posture (p = 0.944). The ANOVA showed no significant effect of posture (p = 0.418) or individual difference (p = 0.959) on the relative measurement error. The normal distribution was followed by the data of standing posture (p = 0.278 in Shapiro-Wilk test) but not those of sitting (p = 0.031) or lying45 • posture (p < 0.001). The Wilcoxon signed-rank test showed no significant difference between the data of sitting and standing (p = 0.496), sitting and lying45 • (p = 0.176), or standing and lying45 • posture (p = 0.889). As shown in Figure 4 , the smallest bias between the SG-derived and reference RRs was from the measurement in sitting posture (bias: −0.0543 bpm, 95% limits of agreement (LoA): −2.952 to 2.843 bpm). The measurement in standing posture had a similar bias but a wider LoA. The measurement in lying45 • posture has the largest bias (−0.501 bpm) and the widest LoA (−8.970 to 7.967 bpm). Thus, the SG measurement was most accurate in sitting posture and least accurate in the lying45 • posture. The linear regression coefficient between RR SG and RR AF measurements was high for the sitting and standing postures (0.9394 and 0.9067, respectively) while the value was lower in the lying45 • posture (0.7744), as shown in Figure 5 . The small regression coefficient in the lying45 • posture was largely related to the result from an outlier (patient #2) due to losing mounting of the IJPT sensor. In this work, the IJPT sensor was comprehensively evaluated on 35 COPD patients to measure the RR at different body postures namely standing, sitting at 90 • , and lying at 45 • positions. The results in this work indicated the high accuracy of the IJPT sensor in RR measurement, with minor differences between different postures. The RR value is one of the most important indicators of several chronic diseases including COPD [50] . RR over 25 bpm is considered as one of the COPD exacerbation signs while the normal range of RR in adults is about 12-20 bpm [40] . RR monitoring of COPD patients is usually carried out in hospitals and clinics via manual counting of breaths, using nasal sensors, or extracted from the electrocardiogram (ECG) signals. The measurement with manual counting is inaccurate [16] while the use of nasal sensors is difficult especially when patients are on continuous oxygen support. Additionally, ECG and nasal sensors are cumbersome and uncomfortable for long-term monitoring. On the other hand, significant progress in the development of WFS has provided the possibility for the convenient, long-term, and low-cost monitoring of vital signs including RR for COPD patients [16, 33] . However, there is a lack of validation and clinical evaluation of WFS on COPD patients in different postures, daily activities, and other physiological conditions where the noises could affect the signal quality [31, 33, 40] . Our results provided a reference for the development of WFS for clinical use. Another challenge is to get accurate estimation of RR in varying breathing patterns. COPD patients usually have an unstable breathing pattern, which causes the variations of RR during measurement. It was observed that RR changed from 15.2 ± 4.3 bpm to 19.1 ± 5.9 bpm during the exacerbation of COPD [24] . COPD patients show characteristic symptoms like breathlessness and cough [51, 52] . These symptoms in addition to body posture are associated with the respiratory volume change. Resultantly, the deformed respiratory waveform and amplitude of peaks can lead to the inaccuracy in RR estimation, as observed in this study. Figure 6 shows a comparison between good and deformed respiratory signals measured by the IJPT. It should be noted that the signal quality depends on many factors including the stability of the sensor attachment on the patients. Another factor that affects the signal is the amount of clothing on the patients. Furthermore, the amount of volume change in the abdominal area varies significantly among individuals and also among postures as reported in the literature [17] . These changes in the volume could significantly affect the ability of the IJPT sensor to measure the RR accurately, as seen in the lying45 • posture, which is the most challenging posture for RR measurement and yields the results consistent with what was reported in the literature [17] . Furthermore, some COPD patients cannot endure standing or sitting for long periods. Therefore, the comprehensive evaluation of RR monitoring in different circumstances is important for the clinical application of the IJPT sensor [30] . The high measurement accuracy of the IJPT sensor at different postures and among different patients highlighted its potential for reliable monitoring of RR in COPD patients. The statistical analysis demonstrated a high accuracy of the IJPT sensor in RR monitoring on COPD patients with good stability at different posture positions. In addition, there was insignificant effect of posture or individual difference on the estimation error. Moreover, the measurement accuracy of the IJPT sensor was comparable with other sensors that have been clinically evaluated on healthy human subjects reported in [35] [36] [37] [38] [39] . As shown in Table 1 , the IJPT sensor showed higher accuracy in RR monitoring on COPD patients (except in the lying posture, which could be related to losing mounting of the IJPT sensor as aforementioned) than most of the existing sensors. Currently, the RR monitoring is often neglected or inaccurately recorded even on patients with respiratory diseases due to the lack of appropriate devices for clinical use [7] . The low fabrication cost, biocompatible substrate, skin-friendly attachment on body surface, and the ease of movement without restrictions in daily activities, as well as the high measurement accuracy of the IJPT sensor make it a promising technology for remote and continuous monitoring of RR for patients with COPD and other respiratory diseases [13] . Reliable RR monitoring for COPD patients using the IJPT sensor could reduce the healthcare cost and the pressure on healthcare facilities especially in low-resource settings such as refugee camps [54] . A recent systematic review disclosed that the majority of physiological monitoring methods for COPD are intermittent with no more than twice a day measurements' frequency [55] . Acute exacerbations of COPD require intensive care treatment immediately. The evaluation of vital signs is necessary to detect physiological abnormalities (micro events), but patients may deteriorate between measurements [56] . The continuous monitoring of RR on COPD patients is an unmet clinical need that has attracted increasing research focus. Especially, the RR monitoring based on WFS on chest band has been proven with the highest reliability compared with other sensors [57] , which makes it possible to provide reliable daily monitoring of COPD based on WFS. Furthermore, the frequency and severity of COPD exacerbation symptoms would be reduced by early detection of abnormal physiological measurements including the respiratory rate [42] . To fight the ongoing pandemic of COVID-19, the combination of internet-of-things (IoT) technology and the IJPT sensor can generate a low-cost, safe, and convenient approach for the remote monitoring, management, and early intervention of the patients where recent studies [58, 59] have mentioned that RR could serve as a leading indicator of COVID-19. There were some limitations associated with this study. Firstly, the IJPT sensor was tested on 35 COPD patients with only one female and also not all of the measurements at the three postures were available due to the limitations aforementioned in Section 2.3. Therefore, further investigation with more diverse COPD patients is required. Secondly, further clinical evaluation of the IJPT sensor on COPD patients during walking, running, sleeping, and other daily activities can be carried out to expand its application scenarios. Finally, longer time series data should be recorded and clinically evaluated on healthy and non-healthy subjects to further investigate the applicability of the IJPT sensor for continuous monitoring of RR. For future work, an innovative mounting mechanism can be developed as the substitute for the fabric belt to achieve more convenient and reliable attachment in different postures. A wireless platform can be designed to achieve more comfortable measurement for the end user, which will ease the remote and continuous monitoring of RR. The results of clinical evaluation on 35 COPD patients in the present work indicated that the IJPT sensor was able to accurately measure the RR at different postures. It can be concluded that the IJPT sensor showed comparable accuracy with other wearable sensors on COPD patients evaluated in the literature with the absolute relative error of 4.49%, 7.29%, and 9.47% at sitting, standing, and lying45 • postures, respectively. The IJPT sensor is promising in achieving reliable RR monitoring for COPD patients where the use of this sensor would contribute to mitigating the frequency and severity of COPD exacerbation symptoms by early detection of abnormal physiological measurements including the respiratory rate. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. The data presented in this study are available in Appendix A. Ethical Statements: All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the frequently always 11 rare sometimes never rare rare frequently frequently sometimes 12 sometimes sometimes never never never frequently always always 13 rare sometimes never never never always always always 14 sometimes never never rare never frequently frequently frequently 15 rare never never never never always always always 16 sometimes frequently frequently rare rare never sometimes frequently 17 rare rare rare never never always frequently frequently 18 frequently rare sometimes never never always always frequently 19 frequently frequently sometimes sometimes rare frequently sometimes frequently 20 frequently frequently rare always always sometimes rare frequently 21 frequently rare always always always always sometimes rare 22 always sometimes sometimes sometimes sometimes always frequently never 23 rare never never never never always frequently frequently 24 sometimes never never never frequently frequently sometimes always 25 sometimes sometimes frequently always always sometimes rare never 26 sometimes always sometimes frequently frequently sometimes sometimes frequently 27 frequently rare never sometimes sometimes always frequently frequently 28 always always rare rare never always always frequently 29 always frequently never always rare frequently rare frequently 30 always always always always frequently rare rare never 31 sometimes never sometimes frequently always rare frequently rare 32 sometimes never never rare rare frequently always always 33 frequently always never sometimes frequently always rare sometimes 34 frequently rare rare sometimes never always sometimes frequently 35 sometimes rare sometimes sometimes rare sometimes frequently sometimes Transfer Entropy Modeling of Newborn Cardiorespiratory Regulation Comparison of different modulations of photoplethysmography in extracting respiratory rate: From a physiological perspective The Importance of Respiratory Rate Monitoring: From Healthcare to Sport and Exercise Accurate respiratory rates count: So should you! Australas. Emerg Time to reconsider how ventilation is regulated above the respiratory compensation point during incremental exercise Monitoring respiratory rate in adults Recent development of respiratory rate measurement technologies Wearable Flexible Sensors: A Review Recent advances in wearable sensors for health monitoring Wearables in Medicine Wearable Medical Devices Using Textile and Flexible Technologies for Ambulatory Monitoring Monitoring of Vital Signs with Flexible and Wearable Medical Devices Flexible, stretchable sensors for wearable health monitoring: Sensing mechanisms, materials, fabrication strategies and features Experimental Characterization of Inkjet-Printed Stretchable Circuits for Wearable Sensor Applications Detecting vital signs with wearablewireless sensors Fabrication and Evaluation of a Novel Non-Invasive Stretchable and Wearable Respiratory Rate Sensor Based on Silver Nanoparticles Using Inkjet Printing Technology Clinical Evaluation of Stretchable and Wearable Inkjet-Printed Strain Gauge Sensor for Respiratory Rate Monitoring at Different Body Postures Respiratory frequency during exercise: The neglected physiological measure Contact-based methods for measuring respiratory rate Smart vest for respiratory rate monitoring of copd patients based on non-contact capacitive sensing A respiration sensor for a chest-strap based wireless body sensor Measurement of heart rate and respiratory rate using a textile-based wearable device in heart failure patients Development of Screen-Printed Breathing Rate Sensors A wearable respiration monitoring system based on digital respiratory inductive plethysmography Wearable Respiratory Rate Monitoring using Piezo-resistive Fabric Sensor The structure design of piezoelectric poly(vinylidene fluoride) (PVDF) polymer-based sensor patch for the respiration monitoring under dynamic walking conditions Clinical applications of wearable technology Mortality in COPD: Causes, Risk Factors, and Prevention Hypnosis for the management of anxiety and dyspnea in COPD: A randomized, sham-controlled crossover trial Monitoring breathing rate at home allows early identification of COPD exacerbations Identification of COPD patients' health status using an intelligent system in the CHRONIOUS wearable platform The importance of respiratory rate monitoring Home monitoring of breathing rate in people with chronic obstructive pulmonary disease: Observational study of feasibility, acceptability, and change after exacerbation Wearable health devices-Vital sign monitoring, systems and technologies Smart Textile Based on Fiber Bragg Grating Sensors for Respiratory Monitoring: Design and Preliminary Trials Wearable radio-frequency sensing of respiratory rate, respiratory volume, and heart rate Monitoring respiratory rates with a wearable system using a stretchable strain sensor during moderate exercise Respiration rate and volume measurements using wearable strain sensors. NPJ Digit Contactless monitoring of breathing patterns and respiratory rate at the pit of the neck: A single camera approach Continuous remote monitoring of COPD patients-justification and explanation of the requirements and a survey of the available technologies Clinical evaluation of stretchable and wearable inkjet-printed strain gauge sensor for respiratory rate monitoring at different measurements locations A Telemedicine-Based Intervention Reduces the Frequency and Severity of COPD Exacerbation Symptoms: A Randomized The revision of the Declaration of Helsinki: Past, present and future e-Health Sensor Platform V2.0 for Arduino and Raspberry Pi [Biometric/Medical Applications Arduino Mega 2560 Datasheet Statistical methods for assessing agreement between two methods of clinical measurement Understanding Bland Altman analysis Optimal assessment and management of chronic obstructive pulmonary disease (COPD) COPD symptoms in the morning: Impact, evaluation and management Observational study to characterise 24-hour COPD symptoms and their relationship with patient-reported outcomes: Results from the ASSESS study Validation of respiratory rate measurements from remote monitoring device in COPD patients Economic impact of remote patient monitoring: An integrated economic model derived from a meta-analysis of randomized controlled trials in heart failure Monitoring of Physiological Parameters to Predict Exacerbations of Chronic Obstructive Pulmonary Disease (COPD): A Systematic Review Physiological abnormalities in patients admitted with acute exacerbation of COPD: An observational study with continuous monitoring Application of Mixed Effects Limits of Agreement in the Presence of Multiple Sources of Variability: Exemplar from the Comparison of Several Devices to Measure Respiratory Rate in COPD Patients Analyzing changes in respiratory rate to predict the risk of COVID-19 infection Rise in nocturnal respiratory rate during CPAP may be an early sign of COVID-19 in patients with obstructive sleep apnea