A lightweight sensing platform for monitoring sleep quality and posture: a simulated validation study

A lightweight sensing platform for monitoring sleep quality and posture: a simulated validation... Background: The prevalence of self-reported shoulder pain in the UK has been estimated at 16%. This has been linked with significant sleep disturbance. It is possible that this relationship is bidirectional, with both symptoms capable of causing the other. Within the field of sleep monitoring, there is a requirement for a mobile and unobtru- sive device capable of monitoring sleep posture and quality. This study investigates the feasibility of a wearable sleep system ( WSS) in accurately detecting sleeping posture and physical activity. Methods: Sixteen healthy subjects were recruited and fitted with three wearable inertial sensors on the trunk and forearms. Ten participants were entered into a ‘Posture’ protocol; assuming a series of common sleeping postures in a simulated bedroom. Five participants completed an ‘Activity’ protocol, in which a triphasic simulated sleep was per- formed including awake, sleep and REM phases. A combined sleep posture and activity protocol was then conducted as a ‘Proof of Concept’ model. Data were used to train a posture detection algorithm, and added to activity to predict sleep phase. Classification accuracy of the WSS was measured during the simulations. Results: The WSS was found to have an overall accuracy of 99.5% in detection of four major postures, and 92.5% in the detection of eight minor postures. Prediction of sleep phase using activity measurements was accurate in 97.3% of the simulations. The ability of the system to accurately detect both posture and activity enabled the design of a conceptual layout for a user-friendly tablet application. Conclusions: The study presents a pervasive wearable sensor platform, which can accurately detect both sleeping posture and activity in non-specialised environments. The extent and accuracy of sleep metrics available advances the current state-of-the-art technology. This has potential diagnostic implications in musculoskeletal pathology and with the addition of alerts may provide therapeutic value in a range of areas including the prevention of pressure sores. Keywords: Shoulder, Pervasive, Monitoring, Posture, Sleep, Activity, Sensors, Wearables Background general practitioners, it is the third most common mus- Pathologies affecting the shoulder are common within culoskeletal presenting complaint [2]. Shoulder pain has the population, often leading to a significant loss of func - been found to lead to significant sleep disturbance, which tion. UK prevalence of self-reported shoulder pain has consequentially has significant impact on quality of life. been estimated at 16%, with rates as high as 26% in the Further, sleeping posture has been implicated as a causa- elderly [1]. Accounting for 2.36% of presentations to tive mechanism of certain shoulder pathologies, as well as having a detrimental impact on post-operative healing following musculoskeletal surgery. Patients following shoulder surgery demonstrated *Correspondence: richard.kwasnicki07@imperial.ac.uk Department of Surgery and Cancer, Imperial College London, 10th Floor greater pain intensity and duration in comparison to total QEQM Building, St. Mary’s Hospital, Praed Street, London W2 1NY, UK hip and knee arthroplasty patient cohorts, with such pain Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 2 of 9 having significantly greater interference with sleep and current work aimed to develop a multi-sensor system activities of daily living [3]. Smith et al. [4] demonstrated architecture combining wearable sensors, communica- that 90% of patients were unable to sleep on their affected tion technology and data analytics in a single platform shoulder post-operatively, additionally 80% of patients capable of ambulatory monitoring. Herein, the presented with shoulder conditions without prior surgery were also study aims to investigate the accuracy of a wearable unable to sleep on their affected shoulder [5]. sleep system (WSS) in the detection of sleep quality and It has been hypothesised that a lateral decubitus sleep- posture. ing position can lead to increased shoulder pressure for extended periods of time, precipitating chronic pain [6]. Methods Kempf et  al. [7] demonstrated a significant correlation Study overview of 68% between the side of shoulder pain and preferred A series of laboratory-based simulations were designed sleeping side, and hypothesised that manipulating sleep- to assess the ability of WSS to detect sleeping posture ing posture could prevent further damage to the shoul- and activity. Healthy adult subjects were recruited locally der. Werner et  al. [8] reported that the supine position at the Hamlyn Centre (Imperial College London) and resulted in significantly lower subacromial pressures excluded if they demonstrated active shoulder pathology compared to prone and lateral decubitus positions, dem- or reported upper body mobility limitation of any sort. onstrating that sleep positions leading to active flexion, Ethical approval was gained from the NRES Committee abduction and internal rotation should be avoided during London—Dulwich on 19th November 2013 (REC Refer- recovery. ence: 10/H0808/124). Informed consent was obtained Monitoring sleep quality may provide broad insight directly from all study participants. into general health status of patients [9–13], with exten- Sleeping posture was investigated by asking healthy sion of monitoring into patients with shoulder patholo- individuals to assume a variety of predefined sleeping gies seeming logical given the high prevalence of sleep positions. Sleep activity was investigated by asking sub- disturbance, anxiety and depression [14]. In the setting jects to simulate a typical sleeping period progressing of shoulder pathology, sleep activity and posture provide through stages of sleep and replicating associated activity surrogate markers for sleep quality and limb position- at each stage. ing. The current gold standard for sleep monitoring is polysomnography (PSG); however, there have been tech- Wearable sleep system (WSS) nological advancements to produce less intrusive meth- A bespoke WSS was designed consisting of a small wear- ods of monitoring sleep, which can be used in a more able sensor positioned on each arm and the chest, (seen natural sleep setting. These include actigraphy (physical in Fig.  1) communicating wirelessly to a local processing activity), heart rate variability (HRV), and smartphone unit (in this study was a laptop computer) using a radiof- applications. Actigraphy and HRV are currently the only requency transceiver. The node platform consists of a TI methods that have been validated to measure sleep qual- MSP430 ultra low power processor, a Chipcon CC2420 ity to a high degree of accuracy, ranging from 82–97 and RF module for wireless communications and a light 57–93%, respectively [15–21]. weight Li-ion polymer battery. The node is integrated Alongside sleep quality, posture monitoring during with an Analog Devices ADXL330 for measurement of sleep provides clinically valuable information regarding 3D acceleration, an InvenSense ITG-3200 digital gyro- the arrangement of limbs and resultant joint angles and scope for 3D angular velocity measurement and a Hon- likely pressures. Electrocardiogram (ECG) waveforms, eywell HMC5843 for 3D magnetic field measurement as pressure sensors and ultra-wide band (UWB) frequency previously described. The whole sensor node measures technology have all previously been used in the deter- 20 × 30 × 17  mm with a weight of 10  g and has previ- mination of sleep posture, though their efficacy in the ously been shown effective in the quantitative analysis of patients with shoulder pain is yet to be established [22– human body movements [35]. 27]. Joint angle measurements provide another dimen- sion to sleep posture monitoring and have demonstrated Posture monitoring efficacy in the orthopaedic patient cohort. The current Participants were first asked to lie on the bed and assume main methodologies in the measurement of joint angles eight postures in a known order, demonstrated in Fig.  2, include inertial sensors, universal goniometer and smart- to allow sensor calibration. Each posture was maintained phone applications, with inertial sensors attracting the for 10  s before participants were instructed to assume greatest interest within the literature [28–34]. However, the next predefined posture. To facilitate an assess - traditional emphasis has been placed upon the refine - ment of blinded classification accuracy, each partici - ment of wearable sensors in isolation. In contrast, the pant re-enacted the postures in a randomised order. The Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 3 of 9 Fig. 1 The WSS sensor platform. a Schematic representation of WSS sensors placement, with one on each arm and one on the trunk. b Structural representation of the WSS sensors used randomised sequence consisted of 16 total postures with movements therefore simulates progression throughout each original posture performed twice by participants. natural sleep. Table  1 outlines the degree of movement A broad range of postures were chosen to determine the permitted in each of the sleep stages A, B and C. The ability of the WSS to recognise both major and minor sleep stage simulation protocol was developed in view movements that may be employed by patients in the clin- of the limited battery life of the WSS (14 min) and was ical setting. In order to direct the selection of such posi- designed to broadly simulate differing stages of sleep tions, focus groups were held with study participants to across the constrained time period. identify common sleeping positions amongst individuals without shoulder pathology. Further sleeping positions in those with shoulder pain were identified by clinical Combined proof of concept model authors through consultation with patients suffering with A single participant completed a simulation includ- shoulder pathologies. ing combination of both posture and activity changes to assess the validity of WSS as a future platform for total sleep monitoring. Sensors were placed as described in Sleep quality monitoring Fig. 1 and the participant was asked to calibrate the sen- Participants were asked to perform a semi-structured sors with known body position as previously described. simulation of the three sleep stages (awake, sleep and Immediately after completion of all eight postures the REM sleep) based upon the degree of major and minor subject was then asked to perform the same three simu- transition movement permitted as outlined in the sleep lated sleep phases. stage simulation protocol, Table  1. Major transitions were defined as movement between the four main pos - Data analysis tures: right slide, left side, supine and prone. Minor Data from tri-axial accelerometer, gyroscope, and mag- transitions were defined as isolated movement of the netometer for all three sensors were integrated and limbs within each of the four main postures. Major and wirelessly sent to a receiving laptop. An algorithm was minor transition movements were standardised before developed to allow the training of a computerised pos- recording commenced through verbal explanation ture classifier based on calibration data (see Additional and the use of a visual aid, Fig.  2. Each sleep stage was file  1). Data were analysed on MATLAB (MathWorks simulated for 2 min with a sequential order of A, B, C, Version R2014a (8.3.0.532) to quantify both individual B, C, B, A, providing 14  min of recording activity for and overall accuracy and error. For activity, a surro- each participant, where A represents awake, B repre- gate for sleep phase, accelerometer values from each of sents sleep, and C represents REM sleep. Our string of Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 4 of 9 Fig. 2 Simulated sleeping postures. (1) Right lateral decubitus—both hands under cheek, arms parallel; (2) right lateral decubitus—bottom forearm under head, top arm relaxed with hand in front of face; (3) supine—arms parallel to body; (4) supine—hands behind head; (5) left lateral decubitus—both hands under cheek, arms parallel; (6) left lateral decubitus—bottom forearm under head, top arm relaxed with hand in front of face; (7) prone—hands in front of forehead; (8) prone—arms parallel to body the three sensors were combined to estimate activity (F3 > 0.14, “3”, IF (F3 > 0.058, “2”, IF (F3 < 0.058, “1”))) level using data variance, i.e. the magnitude of 3-dimen- was used to quantify activity levels and correspond- sional movement. Subject-specific activity thresholds ing sleep stages. To calculate the percentage of time, the were derived from study data and used to determine participant spent in each phase the COUNTIF func- the simulated sleep phase. The nested IF function: = IF tion = COUNTIF(H3:H24509,3) was run across all data. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 5 of 9 Table 1 Sleep stage simulation protocol Simulated Major Minor Movement Actual sleep stage transition transition between transitions permitted permitted permitted Awake Every 30 s No Yes 1 79.6 20.7 00 00 00.37 Sleep No Every minute Limited REM sleep No No No 2 19.6 79.3 00.35000 0 The degree of movement permitted during the each of the simulated sleep 3 00 100 0.70 00 00 stages is outlined along with corresponding major and minor transition frequencies where appropriate 4 00 0 98.2 00 0.34 0.74 Estimate 5 00 00.70 99.7 17.4 00.74 Statistical analysis Statistical tests were used to determine if any of the par- 6 0.77 00 00.32 82.6 00 ticipants or postures demonstrated particularly high and low accuracy levels. The non-parametric test Kruskal– 7 000000 99.7 0 Wallis one-way non-parametric analysis of variance 8 000000 0 98.1 (KW-ANOVA) was used to initially detect any outliers. Fig. 3 Classification matrix representing the percentage accuracy for Outliers were further compared against group averages the eight main postures using the Mann–Whitney U Test. Analyses were per- formed using SPSS version 20.0 for Windows. The statis - tical significance level was set at P < 0.05. of variation calculated from each of the 3 accelerometer axes of each sensor. Results Sixteen healthy subjects were recruited, ten participat- ing in the posture protocol, five in the activity protocol, Discussion and one for the proof on concept simulation. There were This study demonstrates the high accuracy levels achiev - seven females and nine males with a mean age of 25 years able for monitoring physical activity and postures dur- old. No major technical issues arose and datasets were ing sleep with a wearable sensor platform. The WSS was available for all participants recruited. found to have an overall accuracy of 99.5% in the detec- tion of four main postures, which was mostly maintained when detecting eight postures producing an accuracy of Posture 92.5%. Accuracy was acceptable across all subjects with Across 10 participants, the WSS platform showed an the least accurate being 84.3%. The platform could pre - overall classification accuracy of 99.5% for detecting dict simulated sleep phases (awake, sleep, REM) using the four main sleeping postures: right, supine, left, and arm and trunk activity measurements. The ability of the prone. Classification accuracy across all eight postures system to detect both posture and activity was exhibited was 92.5% (Fig. 3). in a proof of concept dataset, along with a conceptual The distinction between postures 1 and 2 was the most layout for a tablet application to be used by both doctors difficult to classify, followed by postures 5 and 6. After and patients (Fig. 5). KW-ANOVA, the mean rank suggested postures 1, 2 and The 4-posture classification accuracy (99.5%) compares 6 to be outliers. A Mann–Whitney U test found no sta- well to other papers in the field. Hsia et  al. [23] used a tistical significance between the classification accuracy Bayesian Classification with pressure sensors finding an of these postures compared to the others. Between par- overall accuracy of 81.4%. The use of a wireless identifica ticipants, the classification accuracy varied from 84.3 to tion and sensing platform by Hoque et  al. [36] gained a 100%. 94.4% accuracy in detection of the four postures, whilst Ni et  al. [27] received similar results to this study in Activity their use of UWB tags combined with a pressure sensor Activity classification across five participants estimated matrix, at approximately 99% accuracy. For eight pos 28.5% of the time spent awake (28.6% simulated), 42.6% tures, the WSS demonstrated an accuracy of 92.5%. Two of the time spent asleep (42.9% simulated) and 26.6% of studies using embedded pressure sensors, one consider- the time spent in REM (28.6% simulated). The activity ing five postures, and the other nine, gained accuracies of measured for the simulated sleep of one participant are 97.7 and 94%, respectively [24, 25]. Although both these shown in Fig.  4, represented as the combined coefficient Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 6 of 9 Fig. 4 Graphical representation of the various stages of sleep quantified using the WSS. a Represents the combined coefficient of variation from all three sensors; b represents the separation of (a) into the three phases: 1 (REM), 2 (sleep) and 3 (awake) papers gained a higher classification accuracy compared are likely to occur, but whether minor differences in posi - to our 92.5%, pressure systems were embedded within tion will affect the utility of the information in currently mattresses, with the system described remaining the unclear. most accurate and advanced wearable platform. The WSS could measure the duration of each sleep Distinction between two right-sided postures (1 and 2) phase with high accuracy, only the REM phase having yielded the greatest identification error during our study. a greater than 1% deviation from the predicted, leading These postures are very similar, with the only difference to an overall accuracy of 97.3%. This validates the ability being 90° rotation at the shoulder in one arm. Therefore, of WSS to assess activity levels of a person whilst they it is possible that the error in these postures is due to are sleeping, facilitating the assessment of sleep qual- poor user compliance with the participant not recreating ity in a natural environment. This compares well to the the posture performed during calibration. These findings actigraph sensor, with the added benefit of simultaneous are replicated on the left side. The possibility of poor par - posture detection. Chang et  al. [37] are the only other ticipant compliance is further highlighted by the fact that group who have been able to provide a platform that is in 2 out of 10 participants the sensors were 100% accu- able to detect both sleeping posture and activity levels. A rate, whilst other participants had accuracy levels as low tri-axial accelerometer was used on the chest for posture as 84.3%. That said, in a true clinical model similar issues detection, combined with ECG recording for sleep stage Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 7 of 9 Fig. 5 Proof of concept output from the pervasive sleep sensor platform presented as part of a conceptual application interface. Reported data include demographics, activity levels with corresponding time intervals and relative posture for utilisation by clinicians and patients monitoring [37]. The method described in this paper pro - such findings in a cohort of patients with shoulder vides additional information regarding the position of the pathology over a natural sleep cycle would prove use- upper limbs which is of particular interest in musculo- ful in determining if similar results can be obtained in skeletal pathology. those with shoulder pain and concurrent sleep distur- The envisaged clinical impact of the WSS is primarily bance. Of note, the sleep postures chosen were designed diagnostic, but with small adjuncts could become therapeu- to capture potential movements replicated in a clini- tic. Widening access of sleep monitoring beyond specialist cal cohort. Future comparative studies will yield valu- facilities would allow patients suffering from upper limb able data regarding actual preferred sleep positions in symptoms to consider if sleep position may be a contribut- patients with shoulder pathology. As a validation study, ing factor. It also facilitates research into the sleep behav- the current work proves useful in providing preliminary iour of post-operative patients, which may give insight into data to inform the design of future comparative studies. why some experience delay in recovery. The addition of an The main technical limitations of the WSS include bat - alert function based on pre-set criteria such as sleeping tery life and sensor size. Unfortunately, the current WSS in one position for too long, might allow for therapeutic battery only lasts for 30  min, making overnight use cur- utility, e.g. following shoulder arthroplasty. In the broader rently unfeasible. To mitigate this, protocols were tailored healthcare setting use of such systems might help prevent accordingly, allowing the simulation of sequences to rep- pressure sores by alerting carers when patients have been in resent part of a night’s sleep. As prototypes, the sensors one position for a certain length of time. are cumbersome (20 × 30 × 17  mm) leading to potential The interpretation and application of the results discomfort. Both limitations could be overcome with should be done so in the context of the study limita- formal sensor design and packaging, and optimising sen- tions. Despite the promising accuracy demonstrated sor settings with regard to frequency of data capture and by the WSS, our data were only collected from a cohort transmission. The study participant demographic was not of healthy patients measured over 14  min of simulated in keeping with that of the clinical cohort; however, this sleep. Future comparative studies seeking to replicate is unlikely to affect the potential utility. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 8 of 9 Conclusions Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- This work demonstrates the accuracy of a wireless sen - lished maps and institutional affiliations. sor platform to detect sleeping posture and quality. This has potential for use in patients with musculoskel - Received: 17 March 2018 Accepted: 18 May 2018 etal pathology, as well as other healthcare applications such as pressure sore prevention. Ultimately, it is hoped that such sensor platforms could provide a low cost, References mobile sleep laboratory, which could facilitate a greater 1. Urwin M, Symmons D, Allison T, Brammah T, Busby H, Roxby M, Simmons number of sleep studies offering insight into disease A, Williams G. Estimating the burden of musculoskeletal disorders in processes, and help tailor the holistic management of the community: the comparative prevalence of symptoms at different anatomical sites, and the relation to social deprivation. Ann Rheum Dis. patients with musculoskeletal disease. 1998;57(11):649–55. 2. Linsell L, Dawson J, Zondervan K, Rose P, Randall T, Fitzpatrick R, Carr A. Additional file Prevalence and incidence of adults consulting for shoulder conditions in UK primary care; patterns of diagnosis and referral. Rheumatology (Oxford). 2006;45(2):215–21. Additional file 1. Posture classification algorithm. (A) The main part of 3. Lindberg MF, Grov EK, Gay CL, Rustoen T, Granheim TI, Amlie E, Lerdal the algorithm that runs in MATLAB used to determine posture classifica- A. Pain characteristics and self-rated health after elective orthopaedic tion. (B) A brief mechanistic overview of the algorithm. surgery—a cross-sectional survey. J Clin Nurs. 2013;22(9–10):1242–53. 4. Smith KL, Harryman DT 2nd, Antoniou J, Campbell B, Sidles JA, Matsen FA 3rd. A prospective, multipractice study of shoulder function and health Abbreviations status in patients with documented rotator cuff tears. J Shoulder Elbow WSS: wearable sleep system; REM: rapid eye movements; PSG: polysomnogra- Surg. 2000;9(5):395–402. phy; UWB: ultra-wide band; HRV: heart rate variability; ECG: echocardiogram; 5. Largacha M, Parsons IT, Campbell B, Titelman RM, Smith KL, Matsen F 3rd. KW-ANOVA: Kruskal–Wallis one-way non-parametric analysis of variance. Deficits in shoulder function and general health associated with sixteen common shoulder diagnoses: a study of 2674 patients. J Shoulder Elbow Authors’ contributions Surg. 2006;15(1):30–9. RK and GC designed the project, collected, and analysed data. RK, LG, and GC 6. Zenian J. Sleep position and shoulder pain. Med Hypotheses. developed the manuscript. ZZ and GZY assisted with software development 2010;74(4):639–43. and provided technical support throughout the duration of the project. PR, 7. Kempf B, Kongsted A. Association between the side of unilateral shoulder AD, and RE provided senior clinical input for the design and conduct of the pain and preferred sleeping position: a cross-sectional study of 83 Danish project. All authors reviewed the manuscript. All authors read and approved patients. J Manipulative Physiol Ther. 2012;35(5):407–12. the final manuscript. 8. Werner CM, Ossendorf C, Meyer DC, Blumenthal S, Gerber C. Subacromial pressures vary with simulated sleep positions. J Shoulder Elbow Surg. Author details 2010;19(7):989–93. Department of Surgery and Cancer, Imperial College London, 10th Floor 9. Baldwin CM, Griffith KA, Nieto FJ, O’Connor GT, Walsleben JA, Redline S. QEQM Building, St. Mary’s Hospital, Praed Street, London W2 1NY, UK. The The association of sleep-disordered breathing and sleep symptoms with Hamlyn Centre, Institute of Global Health Innovation, Imperial College Lon- quality of life in the Sleep Heart Health Study. Sleep. 2001;24(1):96–105. don, London, UK. Imperial College Healthcare NHS Trust, London, UK. 10. Becker HF, Jerrentrup A, Ploch T, Grote L, Penzel T, Sullivan CE, Peter JH. Eec ff t of nasal continuous positive airway pressure treatment on Acknowledgements blood pressure in patients with obstructive sleep apnea. Circulation. The authors would like to thank all volunteers for their time and involvement 2003;107(1):68–73. with the study. 11. Khan A, Latif F, Hawkins B, Tawk M, Sivaram CA, Kinasewitz G. Eec ff ts of obstructive sleep apnea treatment on left atrial volume and left atrial Competing interests volume index. Sleep Breath. 2008;12(2):141–7. The authors certify that they have no affiliations nor financial involvement 12. Punjabi NM, Shahar E, Redline S, Gottlieb DJ, Givelber R, Resnick HE. with any organisation or entity with direct financial interests in the subject Sleep-disordered breathing, glucose intolerance, and insulin resistance: matter of this article. the Sleep Heart Health Study. Am J Epidemiol. 2004;160(6):521–30. 13. Shahar E, Whitney CW, Redline S, Lee ET, Newman AB, Nieto FJ, O’Connor Availability of data and materials GT, Boland LL, Schwartz JE, Samet JM. Sleep-disordered breathing and The datasets used and analysed during the current study are available from cardiovascular disease: cross-sectional results of the Sleep Heart Health the corresponding author on reasonable request. Study. Am J Respir Crit Care Med. 2001;163(1):19–25. 14. Cho CH, Jung SW, Park JY, Song KS, Yu KI. Is shoulder pain for three Consent for publication months or longer correlated with depression, anxiety, and sleep distur- Not applicable. bance? J Shoulder Elbow Surg. 2013;22(2):222–8. 15. Belanger ME, Bernier A, Paquet J, Simard V, Carrier J. Validating actig- Ethics approval and consent to participate raphy as a measure of sleep for preschool children. J Clin Sleep Med. Ethical approval was gained from NRES Committee London—Dulwich, on 2013;9(7):701–6. 19th November 2013, REC Reference: 10/H0808/124. 16. Paquet J, Kawinska A, Carrier J. Wake detection capacity of actigraphy during sleep. Sleep. 2007;30(10):1362–9. Funding 17. Meltzer LJ, Walsh CM, Traylor J, Westin AM. Direct comparison of two new Supported by (1) Engineering and Physical Sciences Research Council actigraphs and polysomnography in children and adolescents. Sleep. (EPSRC)—ESPRIT project; (2) National Institute for Health Research (NIHR) 2012;35(1):159–66. Biomedical Research Centre based at Imperial College Healthcare NHS Trust 18. Dick R, Penzel T, Fietze I, Partinen M, Hein H, Schulz J. AASM standards and Imperial College London. of practice compliant validation of actigraphic sleep analysis from SOMNOwatch versus polysomnographic sleep diagnostics shows high Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 9 of 9 conformity also among subjects with sleep disordered breathing. Physiol In: Donnelly M, Paggetti C, Nugent C, Mokhtari M. Impact analysis of Meas. 2010;31(12):1623–33. solutions for chronic disease prevention and management: 10th inter- 19. Pollak CP, Tryon WW, Nagaraja H, Dzwonczyk R. How accurately does national conference on smart homes and Health Telematics, ICOST 2012, wrist actigraphy identify the states of sleep and wakefulness? Sleep. Artiminio, Italy, June 12–15, 2012 Proceedings. Berlin: Springer Berlin 2001;24(8):957–65. Heidelberg; 2012. p. 132–9. 20. Lewicke AT, Sazonov ES, Schuckers SA. Sleep-wake identification in 28. Gibbs PT, Asada H. Wearable conductive fiber sensors for multi-axis infants: heart rate variability compared to actigraphy. In: Conference human joint angle measurements. J Neuroeng Rehabil. 2005;2(1):7. proceedings: annual international conference of the IEEE engineering in 29. Dejnabadi H, Jolles BM, Aminian K. A new approach to accurate measure- medicine and biology society IEEE engineering in medicine and biology ment of uniaxial joint angles based on a combination of accelerometers society annual conference. 2004. vol. 1, p. 442–5. and gyroscopes. IEEE Trans Biomed Eng. 2005;52(8):1478–84. 21. Ebrahimi F, Setarehdan SK, Ayala-Moyeda J, Nazeran H. Automatic sleep 30. Piotte F, Gravel D, Nadeau S, Moffet H, Bédard C. Reliability of arthrometric staging using empirical mode decomposition, discrete wavelet trans- measurement of shoulder lateral rotation movement in healthy subjects. form, time-domain, and nonlinear dynamics features of heart rate vari- Physiother Theory Pract. 2007;23(3):169–78. ability signals. Comput Methods Programs Biomed. 2013;112(1):47–57. 31. Favre J, Jolles B, Aissaoui R, Aminian K. Ambulatory measurement of 3D 22. Hsia C-C, Hung Y-W, Chiu Y-H, Kang C-H. Bayesian classification for knee joint angle. J Biomech. 2008;41(5):1029–35. bed posture detection based on kurtosis and skewness estimation. In: 32. Piriyaprasarth P, Morris ME, Winter A, Bialocerkowski AE. The reliability of e-health Networking, applications and services, 2008 HealthCom 2008 knee joint position testing using electrogoniometry. BMC Musculoskelet 10th international conference on: 2008: IEEE. 2008. p. 165–8. Disord. 2008;9(1):6. 23. Hsia C, Liou K, Aung A, Foo V, Huang W, Biswas J. Analysis and comparison 33. Cooper G, Sheret I, McMillian L, Siliverdis K, Sha N, Hodgins D, Kenney L, of sleeping posture classification methods using pressure sensitive bed Howard D. Inertial sensor-based knee flexion/extension angle estimation. system. In: Engineering in medicine and biology society, 2009 EMBC 2009 J Biomech. 2009;42(16):2678–85. annual international conference of the IEEE: 2009, IEEE. 2009. p. 6131–4. 34. Jenny JY. Measurement of the knee flexion angle with a Smartphone- 24. Huang W, Wai AAP, Foo SF, Biswas J, Hsia C-C, Liou K. Multimodal sleeping application is precise and accurate. J Arthroplasty. 2013;28(5):784–7. posture classification. In: Pattern recognition (ICPR), 2010 20th interna- 35. Zhang Z-Q, Pansiot J, Lo B, Yang G-Z. Human back movement analysis tional conference on 2010: IEEE. 2010. p. 4336–9. using bsn. In: Body sensor networks (BSN), 2011 international conference 25. Yousefi R, Ostadabbas S, Faezipour M, Farshbaf M, Nourani M, Tamil L, on: 2011, IEEE. 2011. p. 13–8. Pompeo M. Bed posture classification for pressure ulcer prevention. In: 36. Hoque E, Dickerson RF, Stankovic JA. Monitoring body positions and Engineering in medicine and biology society, EMBC, 2011 annual interna- movements during sleep using wisps. In: Wireless Health 2010: ACM; tional conference of the IEEE: 2011, IEEE. 2011. p. 7175–8. 2010. p. 44–53. 26. Schrempf A, Schoßleitner G, Blaha A, Leipold S: Measuring nightly activity, 37. Chang K-M, Liu S-H. Wireless portable electrocardiogram and a tri-axis body weight and body weight change rate with a sensor equipped bed. accelerometer implementation and application on sleep activity moni- In: Engineering in medicine and biology society (EMBC), 2010 annual toring. Telemed E Health. 2011;17(3):177–84. international conference of the IEEE: 2010, IEEE. 2010. p. 2151–4. 27. Ni H, Abdulrazak B, Zhang D, Wu S, Zhou X, Miao K, Han D. Multi-modal non-intrusive sleep pattern recognition in elder assistive environment. Ready to submit your research ? Choose BMC and benefit from: fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png European Journal of Medical Research Springer Journals

A lightweight sensing platform for monitoring sleep quality and posture: a simulated validation study

Free
9 pages

Loading next page...
 
/lp/springer_journal/a-lightweight-sensing-platform-for-monitoring-sleep-quality-and-k1fiBgs8oY
Publisher
Springer Journals
Copyright
Copyright © 2018 by The Author(s)
Subject
Medicine & Public Health; Medicine/Public Health, general; Infectious Diseases; Internal Medicine; Surgery; Oncology; Biomedicine, general
eISSN
2047-783X
D.O.I.
10.1186/s40001-018-0326-9
Publisher site
See Article on Publisher Site

Abstract

Background: The prevalence of self-reported shoulder pain in the UK has been estimated at 16%. This has been linked with significant sleep disturbance. It is possible that this relationship is bidirectional, with both symptoms capable of causing the other. Within the field of sleep monitoring, there is a requirement for a mobile and unobtru- sive device capable of monitoring sleep posture and quality. This study investigates the feasibility of a wearable sleep system ( WSS) in accurately detecting sleeping posture and physical activity. Methods: Sixteen healthy subjects were recruited and fitted with three wearable inertial sensors on the trunk and forearms. Ten participants were entered into a ‘Posture’ protocol; assuming a series of common sleeping postures in a simulated bedroom. Five participants completed an ‘Activity’ protocol, in which a triphasic simulated sleep was per- formed including awake, sleep and REM phases. A combined sleep posture and activity protocol was then conducted as a ‘Proof of Concept’ model. Data were used to train a posture detection algorithm, and added to activity to predict sleep phase. Classification accuracy of the WSS was measured during the simulations. Results: The WSS was found to have an overall accuracy of 99.5% in detection of four major postures, and 92.5% in the detection of eight minor postures. Prediction of sleep phase using activity measurements was accurate in 97.3% of the simulations. The ability of the system to accurately detect both posture and activity enabled the design of a conceptual layout for a user-friendly tablet application. Conclusions: The study presents a pervasive wearable sensor platform, which can accurately detect both sleeping posture and activity in non-specialised environments. The extent and accuracy of sleep metrics available advances the current state-of-the-art technology. This has potential diagnostic implications in musculoskeletal pathology and with the addition of alerts may provide therapeutic value in a range of areas including the prevention of pressure sores. Keywords: Shoulder, Pervasive, Monitoring, Posture, Sleep, Activity, Sensors, Wearables Background general practitioners, it is the third most common mus- Pathologies affecting the shoulder are common within culoskeletal presenting complaint [2]. Shoulder pain has the population, often leading to a significant loss of func - been found to lead to significant sleep disturbance, which tion. UK prevalence of self-reported shoulder pain has consequentially has significant impact on quality of life. been estimated at 16%, with rates as high as 26% in the Further, sleeping posture has been implicated as a causa- elderly [1]. Accounting for 2.36% of presentations to tive mechanism of certain shoulder pathologies, as well as having a detrimental impact on post-operative healing following musculoskeletal surgery. Patients following shoulder surgery demonstrated *Correspondence: richard.kwasnicki07@imperial.ac.uk Department of Surgery and Cancer, Imperial College London, 10th Floor greater pain intensity and duration in comparison to total QEQM Building, St. Mary’s Hospital, Praed Street, London W2 1NY, UK hip and knee arthroplasty patient cohorts, with such pain Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creat iveco mmons .org/ publi cdoma in/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 2 of 9 having significantly greater interference with sleep and current work aimed to develop a multi-sensor system activities of daily living [3]. Smith et al. [4] demonstrated architecture combining wearable sensors, communica- that 90% of patients were unable to sleep on their affected tion technology and data analytics in a single platform shoulder post-operatively, additionally 80% of patients capable of ambulatory monitoring. Herein, the presented with shoulder conditions without prior surgery were also study aims to investigate the accuracy of a wearable unable to sleep on their affected shoulder [5]. sleep system (WSS) in the detection of sleep quality and It has been hypothesised that a lateral decubitus sleep- posture. ing position can lead to increased shoulder pressure for extended periods of time, precipitating chronic pain [6]. Methods Kempf et  al. [7] demonstrated a significant correlation Study overview of 68% between the side of shoulder pain and preferred A series of laboratory-based simulations were designed sleeping side, and hypothesised that manipulating sleep- to assess the ability of WSS to detect sleeping posture ing posture could prevent further damage to the shoul- and activity. Healthy adult subjects were recruited locally der. Werner et  al. [8] reported that the supine position at the Hamlyn Centre (Imperial College London) and resulted in significantly lower subacromial pressures excluded if they demonstrated active shoulder pathology compared to prone and lateral decubitus positions, dem- or reported upper body mobility limitation of any sort. onstrating that sleep positions leading to active flexion, Ethical approval was gained from the NRES Committee abduction and internal rotation should be avoided during London—Dulwich on 19th November 2013 (REC Refer- recovery. ence: 10/H0808/124). Informed consent was obtained Monitoring sleep quality may provide broad insight directly from all study participants. into general health status of patients [9–13], with exten- Sleeping posture was investigated by asking healthy sion of monitoring into patients with shoulder patholo- individuals to assume a variety of predefined sleeping gies seeming logical given the high prevalence of sleep positions. Sleep activity was investigated by asking sub- disturbance, anxiety and depression [14]. In the setting jects to simulate a typical sleeping period progressing of shoulder pathology, sleep activity and posture provide through stages of sleep and replicating associated activity surrogate markers for sleep quality and limb position- at each stage. ing. The current gold standard for sleep monitoring is polysomnography (PSG); however, there have been tech- Wearable sleep system (WSS) nological advancements to produce less intrusive meth- A bespoke WSS was designed consisting of a small wear- ods of monitoring sleep, which can be used in a more able sensor positioned on each arm and the chest, (seen natural sleep setting. These include actigraphy (physical in Fig.  1) communicating wirelessly to a local processing activity), heart rate variability (HRV), and smartphone unit (in this study was a laptop computer) using a radiof- applications. Actigraphy and HRV are currently the only requency transceiver. The node platform consists of a TI methods that have been validated to measure sleep qual- MSP430 ultra low power processor, a Chipcon CC2420 ity to a high degree of accuracy, ranging from 82–97 and RF module for wireless communications and a light 57–93%, respectively [15–21]. weight Li-ion polymer battery. The node is integrated Alongside sleep quality, posture monitoring during with an Analog Devices ADXL330 for measurement of sleep provides clinically valuable information regarding 3D acceleration, an InvenSense ITG-3200 digital gyro- the arrangement of limbs and resultant joint angles and scope for 3D angular velocity measurement and a Hon- likely pressures. Electrocardiogram (ECG) waveforms, eywell HMC5843 for 3D magnetic field measurement as pressure sensors and ultra-wide band (UWB) frequency previously described. The whole sensor node measures technology have all previously been used in the deter- 20 × 30 × 17  mm with a weight of 10  g and has previ- mination of sleep posture, though their efficacy in the ously been shown effective in the quantitative analysis of patients with shoulder pain is yet to be established [22– human body movements [35]. 27]. Joint angle measurements provide another dimen- sion to sleep posture monitoring and have demonstrated Posture monitoring efficacy in the orthopaedic patient cohort. The current Participants were first asked to lie on the bed and assume main methodologies in the measurement of joint angles eight postures in a known order, demonstrated in Fig.  2, include inertial sensors, universal goniometer and smart- to allow sensor calibration. Each posture was maintained phone applications, with inertial sensors attracting the for 10  s before participants were instructed to assume greatest interest within the literature [28–34]. However, the next predefined posture. To facilitate an assess - traditional emphasis has been placed upon the refine - ment of blinded classification accuracy, each partici - ment of wearable sensors in isolation. In contrast, the pant re-enacted the postures in a randomised order. The Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 3 of 9 Fig. 1 The WSS sensor platform. a Schematic representation of WSS sensors placement, with one on each arm and one on the trunk. b Structural representation of the WSS sensors used randomised sequence consisted of 16 total postures with movements therefore simulates progression throughout each original posture performed twice by participants. natural sleep. Table  1 outlines the degree of movement A broad range of postures were chosen to determine the permitted in each of the sleep stages A, B and C. The ability of the WSS to recognise both major and minor sleep stage simulation protocol was developed in view movements that may be employed by patients in the clin- of the limited battery life of the WSS (14 min) and was ical setting. In order to direct the selection of such posi- designed to broadly simulate differing stages of sleep tions, focus groups were held with study participants to across the constrained time period. identify common sleeping positions amongst individuals without shoulder pathology. Further sleeping positions in those with shoulder pain were identified by clinical Combined proof of concept model authors through consultation with patients suffering with A single participant completed a simulation includ- shoulder pathologies. ing combination of both posture and activity changes to assess the validity of WSS as a future platform for total sleep monitoring. Sensors were placed as described in Sleep quality monitoring Fig. 1 and the participant was asked to calibrate the sen- Participants were asked to perform a semi-structured sors with known body position as previously described. simulation of the three sleep stages (awake, sleep and Immediately after completion of all eight postures the REM sleep) based upon the degree of major and minor subject was then asked to perform the same three simu- transition movement permitted as outlined in the sleep lated sleep phases. stage simulation protocol, Table  1. Major transitions were defined as movement between the four main pos - Data analysis tures: right slide, left side, supine and prone. Minor Data from tri-axial accelerometer, gyroscope, and mag- transitions were defined as isolated movement of the netometer for all three sensors were integrated and limbs within each of the four main postures. Major and wirelessly sent to a receiving laptop. An algorithm was minor transition movements were standardised before developed to allow the training of a computerised pos- recording commenced through verbal explanation ture classifier based on calibration data (see Additional and the use of a visual aid, Fig.  2. Each sleep stage was file  1). Data were analysed on MATLAB (MathWorks simulated for 2 min with a sequential order of A, B, C, Version R2014a (8.3.0.532) to quantify both individual B, C, B, A, providing 14  min of recording activity for and overall accuracy and error. For activity, a surro- each participant, where A represents awake, B repre- gate for sleep phase, accelerometer values from each of sents sleep, and C represents REM sleep. Our string of Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 4 of 9 Fig. 2 Simulated sleeping postures. (1) Right lateral decubitus—both hands under cheek, arms parallel; (2) right lateral decubitus—bottom forearm under head, top arm relaxed with hand in front of face; (3) supine—arms parallel to body; (4) supine—hands behind head; (5) left lateral decubitus—both hands under cheek, arms parallel; (6) left lateral decubitus—bottom forearm under head, top arm relaxed with hand in front of face; (7) prone—hands in front of forehead; (8) prone—arms parallel to body the three sensors were combined to estimate activity (F3 > 0.14, “3”, IF (F3 > 0.058, “2”, IF (F3 < 0.058, “1”))) level using data variance, i.e. the magnitude of 3-dimen- was used to quantify activity levels and correspond- sional movement. Subject-specific activity thresholds ing sleep stages. To calculate the percentage of time, the were derived from study data and used to determine participant spent in each phase the COUNTIF func- the simulated sleep phase. The nested IF function: = IF tion = COUNTIF(H3:H24509,3) was run across all data. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 5 of 9 Table 1 Sleep stage simulation protocol Simulated Major Minor Movement Actual sleep stage transition transition between transitions permitted permitted permitted Awake Every 30 s No Yes 1 79.6 20.7 00 00 00.37 Sleep No Every minute Limited REM sleep No No No 2 19.6 79.3 00.35000 0 The degree of movement permitted during the each of the simulated sleep 3 00 100 0.70 00 00 stages is outlined along with corresponding major and minor transition frequencies where appropriate 4 00 0 98.2 00 0.34 0.74 Estimate 5 00 00.70 99.7 17.4 00.74 Statistical analysis Statistical tests were used to determine if any of the par- 6 0.77 00 00.32 82.6 00 ticipants or postures demonstrated particularly high and low accuracy levels. The non-parametric test Kruskal– 7 000000 99.7 0 Wallis one-way non-parametric analysis of variance 8 000000 0 98.1 (KW-ANOVA) was used to initially detect any outliers. Fig. 3 Classification matrix representing the percentage accuracy for Outliers were further compared against group averages the eight main postures using the Mann–Whitney U Test. Analyses were per- formed using SPSS version 20.0 for Windows. The statis - tical significance level was set at P < 0.05. of variation calculated from each of the 3 accelerometer axes of each sensor. Results Sixteen healthy subjects were recruited, ten participat- ing in the posture protocol, five in the activity protocol, Discussion and one for the proof on concept simulation. There were This study demonstrates the high accuracy levels achiev - seven females and nine males with a mean age of 25 years able for monitoring physical activity and postures dur- old. No major technical issues arose and datasets were ing sleep with a wearable sensor platform. The WSS was available for all participants recruited. found to have an overall accuracy of 99.5% in the detec- tion of four main postures, which was mostly maintained when detecting eight postures producing an accuracy of Posture 92.5%. Accuracy was acceptable across all subjects with Across 10 participants, the WSS platform showed an the least accurate being 84.3%. The platform could pre - overall classification accuracy of 99.5% for detecting dict simulated sleep phases (awake, sleep, REM) using the four main sleeping postures: right, supine, left, and arm and trunk activity measurements. The ability of the prone. Classification accuracy across all eight postures system to detect both posture and activity was exhibited was 92.5% (Fig. 3). in a proof of concept dataset, along with a conceptual The distinction between postures 1 and 2 was the most layout for a tablet application to be used by both doctors difficult to classify, followed by postures 5 and 6. After and patients (Fig. 5). KW-ANOVA, the mean rank suggested postures 1, 2 and The 4-posture classification accuracy (99.5%) compares 6 to be outliers. A Mann–Whitney U test found no sta- well to other papers in the field. Hsia et  al. [23] used a tistical significance between the classification accuracy Bayesian Classification with pressure sensors finding an of these postures compared to the others. Between par- overall accuracy of 81.4%. The use of a wireless identifica ticipants, the classification accuracy varied from 84.3 to tion and sensing platform by Hoque et  al. [36] gained a 100%. 94.4% accuracy in detection of the four postures, whilst Ni et  al. [27] received similar results to this study in Activity their use of UWB tags combined with a pressure sensor Activity classification across five participants estimated matrix, at approximately 99% accuracy. For eight pos 28.5% of the time spent awake (28.6% simulated), 42.6% tures, the WSS demonstrated an accuracy of 92.5%. Two of the time spent asleep (42.9% simulated) and 26.6% of studies using embedded pressure sensors, one consider- the time spent in REM (28.6% simulated). The activity ing five postures, and the other nine, gained accuracies of measured for the simulated sleep of one participant are 97.7 and 94%, respectively [24, 25]. Although both these shown in Fig.  4, represented as the combined coefficient Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 6 of 9 Fig. 4 Graphical representation of the various stages of sleep quantified using the WSS. a Represents the combined coefficient of variation from all three sensors; b represents the separation of (a) into the three phases: 1 (REM), 2 (sleep) and 3 (awake) papers gained a higher classification accuracy compared are likely to occur, but whether minor differences in posi - to our 92.5%, pressure systems were embedded within tion will affect the utility of the information in currently mattresses, with the system described remaining the unclear. most accurate and advanced wearable platform. The WSS could measure the duration of each sleep Distinction between two right-sided postures (1 and 2) phase with high accuracy, only the REM phase having yielded the greatest identification error during our study. a greater than 1% deviation from the predicted, leading These postures are very similar, with the only difference to an overall accuracy of 97.3%. This validates the ability being 90° rotation at the shoulder in one arm. Therefore, of WSS to assess activity levels of a person whilst they it is possible that the error in these postures is due to are sleeping, facilitating the assessment of sleep qual- poor user compliance with the participant not recreating ity in a natural environment. This compares well to the the posture performed during calibration. These findings actigraph sensor, with the added benefit of simultaneous are replicated on the left side. The possibility of poor par - posture detection. Chang et  al. [37] are the only other ticipant compliance is further highlighted by the fact that group who have been able to provide a platform that is in 2 out of 10 participants the sensors were 100% accu- able to detect both sleeping posture and activity levels. A rate, whilst other participants had accuracy levels as low tri-axial accelerometer was used on the chest for posture as 84.3%. That said, in a true clinical model similar issues detection, combined with ECG recording for sleep stage Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 7 of 9 Fig. 5 Proof of concept output from the pervasive sleep sensor platform presented as part of a conceptual application interface. Reported data include demographics, activity levels with corresponding time intervals and relative posture for utilisation by clinicians and patients monitoring [37]. The method described in this paper pro - such findings in a cohort of patients with shoulder vides additional information regarding the position of the pathology over a natural sleep cycle would prove use- upper limbs which is of particular interest in musculo- ful in determining if similar results can be obtained in skeletal pathology. those with shoulder pain and concurrent sleep distur- The envisaged clinical impact of the WSS is primarily bance. Of note, the sleep postures chosen were designed diagnostic, but with small adjuncts could become therapeu- to capture potential movements replicated in a clini- tic. Widening access of sleep monitoring beyond specialist cal cohort. Future comparative studies will yield valu- facilities would allow patients suffering from upper limb able data regarding actual preferred sleep positions in symptoms to consider if sleep position may be a contribut- patients with shoulder pathology. As a validation study, ing factor. It also facilitates research into the sleep behav- the current work proves useful in providing preliminary iour of post-operative patients, which may give insight into data to inform the design of future comparative studies. why some experience delay in recovery. The addition of an The main technical limitations of the WSS include bat - alert function based on pre-set criteria such as sleeping tery life and sensor size. Unfortunately, the current WSS in one position for too long, might allow for therapeutic battery only lasts for 30  min, making overnight use cur- utility, e.g. following shoulder arthroplasty. In the broader rently unfeasible. To mitigate this, protocols were tailored healthcare setting use of such systems might help prevent accordingly, allowing the simulation of sequences to rep- pressure sores by alerting carers when patients have been in resent part of a night’s sleep. As prototypes, the sensors one position for a certain length of time. are cumbersome (20 × 30 × 17  mm) leading to potential The interpretation and application of the results discomfort. Both limitations could be overcome with should be done so in the context of the study limita- formal sensor design and packaging, and optimising sen- tions. Despite the promising accuracy demonstrated sor settings with regard to frequency of data capture and by the WSS, our data were only collected from a cohort transmission. The study participant demographic was not of healthy patients measured over 14  min of simulated in keeping with that of the clinical cohort; however, this sleep. Future comparative studies seeking to replicate is unlikely to affect the potential utility. Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 8 of 9 Conclusions Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- This work demonstrates the accuracy of a wireless sen - lished maps and institutional affiliations. sor platform to detect sleeping posture and quality. This has potential for use in patients with musculoskel - Received: 17 March 2018 Accepted: 18 May 2018 etal pathology, as well as other healthcare applications such as pressure sore prevention. Ultimately, it is hoped that such sensor platforms could provide a low cost, References mobile sleep laboratory, which could facilitate a greater 1. Urwin M, Symmons D, Allison T, Brammah T, Busby H, Roxby M, Simmons number of sleep studies offering insight into disease A, Williams G. Estimating the burden of musculoskeletal disorders in processes, and help tailor the holistic management of the community: the comparative prevalence of symptoms at different anatomical sites, and the relation to social deprivation. Ann Rheum Dis. patients with musculoskeletal disease. 1998;57(11):649–55. 2. Linsell L, Dawson J, Zondervan K, Rose P, Randall T, Fitzpatrick R, Carr A. Additional file Prevalence and incidence of adults consulting for shoulder conditions in UK primary care; patterns of diagnosis and referral. Rheumatology (Oxford). 2006;45(2):215–21. Additional file 1. Posture classification algorithm. (A) The main part of 3. Lindberg MF, Grov EK, Gay CL, Rustoen T, Granheim TI, Amlie E, Lerdal the algorithm that runs in MATLAB used to determine posture classifica- A. Pain characteristics and self-rated health after elective orthopaedic tion. (B) A brief mechanistic overview of the algorithm. surgery—a cross-sectional survey. J Clin Nurs. 2013;22(9–10):1242–53. 4. Smith KL, Harryman DT 2nd, Antoniou J, Campbell B, Sidles JA, Matsen FA 3rd. A prospective, multipractice study of shoulder function and health Abbreviations status in patients with documented rotator cuff tears. J Shoulder Elbow WSS: wearable sleep system; REM: rapid eye movements; PSG: polysomnogra- Surg. 2000;9(5):395–402. phy; UWB: ultra-wide band; HRV: heart rate variability; ECG: echocardiogram; 5. Largacha M, Parsons IT, Campbell B, Titelman RM, Smith KL, Matsen F 3rd. KW-ANOVA: Kruskal–Wallis one-way non-parametric analysis of variance. Deficits in shoulder function and general health associated with sixteen common shoulder diagnoses: a study of 2674 patients. J Shoulder Elbow Authors’ contributions Surg. 2006;15(1):30–9. RK and GC designed the project, collected, and analysed data. RK, LG, and GC 6. Zenian J. Sleep position and shoulder pain. Med Hypotheses. developed the manuscript. ZZ and GZY assisted with software development 2010;74(4):639–43. and provided technical support throughout the duration of the project. PR, 7. Kempf B, Kongsted A. Association between the side of unilateral shoulder AD, and RE provided senior clinical input for the design and conduct of the pain and preferred sleeping position: a cross-sectional study of 83 Danish project. All authors reviewed the manuscript. All authors read and approved patients. J Manipulative Physiol Ther. 2012;35(5):407–12. the final manuscript. 8. Werner CM, Ossendorf C, Meyer DC, Blumenthal S, Gerber C. Subacromial pressures vary with simulated sleep positions. J Shoulder Elbow Surg. Author details 2010;19(7):989–93. Department of Surgery and Cancer, Imperial College London, 10th Floor 9. Baldwin CM, Griffith KA, Nieto FJ, O’Connor GT, Walsleben JA, Redline S. QEQM Building, St. Mary’s Hospital, Praed Street, London W2 1NY, UK. The The association of sleep-disordered breathing and sleep symptoms with Hamlyn Centre, Institute of Global Health Innovation, Imperial College Lon- quality of life in the Sleep Heart Health Study. Sleep. 2001;24(1):96–105. don, London, UK. Imperial College Healthcare NHS Trust, London, UK. 10. Becker HF, Jerrentrup A, Ploch T, Grote L, Penzel T, Sullivan CE, Peter JH. Eec ff t of nasal continuous positive airway pressure treatment on Acknowledgements blood pressure in patients with obstructive sleep apnea. Circulation. The authors would like to thank all volunteers for their time and involvement 2003;107(1):68–73. with the study. 11. Khan A, Latif F, Hawkins B, Tawk M, Sivaram CA, Kinasewitz G. Eec ff ts of obstructive sleep apnea treatment on left atrial volume and left atrial Competing interests volume index. Sleep Breath. 2008;12(2):141–7. The authors certify that they have no affiliations nor financial involvement 12. Punjabi NM, Shahar E, Redline S, Gottlieb DJ, Givelber R, Resnick HE. with any organisation or entity with direct financial interests in the subject Sleep-disordered breathing, glucose intolerance, and insulin resistance: matter of this article. the Sleep Heart Health Study. Am J Epidemiol. 2004;160(6):521–30. 13. Shahar E, Whitney CW, Redline S, Lee ET, Newman AB, Nieto FJ, O’Connor Availability of data and materials GT, Boland LL, Schwartz JE, Samet JM. Sleep-disordered breathing and The datasets used and analysed during the current study are available from cardiovascular disease: cross-sectional results of the Sleep Heart Health the corresponding author on reasonable request. Study. Am J Respir Crit Care Med. 2001;163(1):19–25. 14. Cho CH, Jung SW, Park JY, Song KS, Yu KI. Is shoulder pain for three Consent for publication months or longer correlated with depression, anxiety, and sleep distur- Not applicable. bance? J Shoulder Elbow Surg. 2013;22(2):222–8. 15. Belanger ME, Bernier A, Paquet J, Simard V, Carrier J. Validating actig- Ethics approval and consent to participate raphy as a measure of sleep for preschool children. J Clin Sleep Med. Ethical approval was gained from NRES Committee London—Dulwich, on 2013;9(7):701–6. 19th November 2013, REC Reference: 10/H0808/124. 16. Paquet J, Kawinska A, Carrier J. Wake detection capacity of actigraphy during sleep. Sleep. 2007;30(10):1362–9. Funding 17. Meltzer LJ, Walsh CM, Traylor J, Westin AM. Direct comparison of two new Supported by (1) Engineering and Physical Sciences Research Council actigraphs and polysomnography in children and adolescents. Sleep. (EPSRC)—ESPRIT project; (2) National Institute for Health Research (NIHR) 2012;35(1):159–66. Biomedical Research Centre based at Imperial College Healthcare NHS Trust 18. Dick R, Penzel T, Fietze I, Partinen M, Hein H, Schulz J. AASM standards and Imperial College London. of practice compliant validation of actigraphic sleep analysis from SOMNOwatch versus polysomnographic sleep diagnostics shows high Kwasnicki et al. Eur J Med Res (2018) 23:28 Page 9 of 9 conformity also among subjects with sleep disordered breathing. Physiol In: Donnelly M, Paggetti C, Nugent C, Mokhtari M. Impact analysis of Meas. 2010;31(12):1623–33. solutions for chronic disease prevention and management: 10th inter- 19. Pollak CP, Tryon WW, Nagaraja H, Dzwonczyk R. How accurately does national conference on smart homes and Health Telematics, ICOST 2012, wrist actigraphy identify the states of sleep and wakefulness? Sleep. Artiminio, Italy, June 12–15, 2012 Proceedings. Berlin: Springer Berlin 2001;24(8):957–65. Heidelberg; 2012. p. 132–9. 20. Lewicke AT, Sazonov ES, Schuckers SA. Sleep-wake identification in 28. Gibbs PT, Asada H. Wearable conductive fiber sensors for multi-axis infants: heart rate variability compared to actigraphy. In: Conference human joint angle measurements. J Neuroeng Rehabil. 2005;2(1):7. proceedings: annual international conference of the IEEE engineering in 29. Dejnabadi H, Jolles BM, Aminian K. A new approach to accurate measure- medicine and biology society IEEE engineering in medicine and biology ment of uniaxial joint angles based on a combination of accelerometers society annual conference. 2004. vol. 1, p. 442–5. and gyroscopes. IEEE Trans Biomed Eng. 2005;52(8):1478–84. 21. Ebrahimi F, Setarehdan SK, Ayala-Moyeda J, Nazeran H. Automatic sleep 30. Piotte F, Gravel D, Nadeau S, Moffet H, Bédard C. Reliability of arthrometric staging using empirical mode decomposition, discrete wavelet trans- measurement of shoulder lateral rotation movement in healthy subjects. form, time-domain, and nonlinear dynamics features of heart rate vari- Physiother Theory Pract. 2007;23(3):169–78. ability signals. Comput Methods Programs Biomed. 2013;112(1):47–57. 31. Favre J, Jolles B, Aissaoui R, Aminian K. Ambulatory measurement of 3D 22. Hsia C-C, Hung Y-W, Chiu Y-H, Kang C-H. Bayesian classification for knee joint angle. J Biomech. 2008;41(5):1029–35. bed posture detection based on kurtosis and skewness estimation. In: 32. Piriyaprasarth P, Morris ME, Winter A, Bialocerkowski AE. The reliability of e-health Networking, applications and services, 2008 HealthCom 2008 knee joint position testing using electrogoniometry. BMC Musculoskelet 10th international conference on: 2008: IEEE. 2008. p. 165–8. Disord. 2008;9(1):6. 23. Hsia C, Liou K, Aung A, Foo V, Huang W, Biswas J. Analysis and comparison 33. Cooper G, Sheret I, McMillian L, Siliverdis K, Sha N, Hodgins D, Kenney L, of sleeping posture classification methods using pressure sensitive bed Howard D. Inertial sensor-based knee flexion/extension angle estimation. system. In: Engineering in medicine and biology society, 2009 EMBC 2009 J Biomech. 2009;42(16):2678–85. annual international conference of the IEEE: 2009, IEEE. 2009. p. 6131–4. 34. Jenny JY. Measurement of the knee flexion angle with a Smartphone- 24. Huang W, Wai AAP, Foo SF, Biswas J, Hsia C-C, Liou K. Multimodal sleeping application is precise and accurate. J Arthroplasty. 2013;28(5):784–7. posture classification. In: Pattern recognition (ICPR), 2010 20th interna- 35. Zhang Z-Q, Pansiot J, Lo B, Yang G-Z. Human back movement analysis tional conference on 2010: IEEE. 2010. p. 4336–9. using bsn. In: Body sensor networks (BSN), 2011 international conference 25. Yousefi R, Ostadabbas S, Faezipour M, Farshbaf M, Nourani M, Tamil L, on: 2011, IEEE. 2011. p. 13–8. Pompeo M. Bed posture classification for pressure ulcer prevention. In: 36. Hoque E, Dickerson RF, Stankovic JA. Monitoring body positions and Engineering in medicine and biology society, EMBC, 2011 annual interna- movements during sleep using wisps. In: Wireless Health 2010: ACM; tional conference of the IEEE: 2011, IEEE. 2011. p. 7175–8. 2010. p. 44–53. 26. Schrempf A, Schoßleitner G, Blaha A, Leipold S: Measuring nightly activity, 37. Chang K-M, Liu S-H. Wireless portable electrocardiogram and a tri-axis body weight and body weight change rate with a sensor equipped bed. accelerometer implementation and application on sleep activity moni- In: Engineering in medicine and biology society (EMBC), 2010 annual toring. Telemed E Health. 2011;17(3):177–84. international conference of the IEEE: 2010, IEEE. 2010. p. 2151–4. 27. Ni H, Abdulrazak B, Zhang D, Wu S, Zhou X, Miao K, Han D. Multi-modal non-intrusive sleep pattern recognition in elder assistive environment. Ready to submit your research ? Choose BMC and benefit from: fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

European Journal of Medical ResearchSpringer Journals

Published: May 30, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off