Journal of Clinical Monitoring and Computing

https://doi.org/10.1007/s10877-019-00329-5

ORIGINAL RESEARCH

Non‑invasive continuous respiratory monitoring using

temperature‑based sensors
Daniel E. Hurtado1   · Angel Abusleme2 · Javier A. P. Chávez3

Received: 7 December 2018 / Accepted: 29 May 2019

© Springer Nature B.V. 2019

Abstract
Respiratory rate (RR) is a key vital sign that has been traditionally employed in the clinical assessment of patients and in the
prevention of respiratory compromise. Despite its relevance, current practice for monitoring RR in non-intubated patients
strongly relies on visual counting, which delivers an intermittent and error-prone assessment of the respiratory status. Here,
we present a novel non-invasive respiratory monitor that continuously measures the RR in human subjects. The respiratory
activity of the user is inferred by sensing the thermal transfer between the breathing airflow and a temperature sensor placed
between the nose and the mouth. The performance of the respiratory monitor is assessed through respiratory experiments
performed on healthy subjects. Under spontaneous breathing, the mean RR difference between our respiratory monitor and
visual counting was 0.4 breaths per minute (BPM), with a 95% confidence interval equal to [− 0.5, 1.3] BPM. The robustness
of the respiratory sensor to the position is assessed by studying the signal-to-noise ratio in different locations on the upper
lip, displaying a markedly better performance than traditional thermal sensors used for respiratory airflow measurements.

Keywords  Monitoring · Respiration · Thermal sensors

1 Introduction that can occur in patients undergoing pain-management

treatments based on opioids who present low RR [5, 12].
Continuous non-invasive respiratory monitoring has gained The prevention of respiratory compromise has been recog-
renewed interest in hospital settings such as the post-anes- nized as a priority by the Joint Commission, highlighting the
thesia care unit (PACU), the intensive care unit, and the relevance of developing reliable and cost-effective respira-
general floor, as it can provide early warnings and alerts of tory monitoring systems for RR [15].
respiratory compromise in non-intubated patients [4, 21]. Despite the importance of continuously monitoring RR in
Respiratory rate (RR) is a key vital sign that has long been hospital patients, the current standard of care for assessing
used as a primary parameter in the evaluation of the health the respiratory condition largely relies on visual counting,
condition of a patient [11]. In particular, RR plays a key role which leads to an intermittent and inaccurate measurement
in the prevention and management of postoperative opioid- of RR [10]. Pulse oximeters have been traditionally used to
induced respiratory depression, a life-threatening condition assess the respiratory condition of patients, and have gained
their popularity based on their non-invasiveness, cost-effec-
tiveness and worldwide availability [6]. However, pulse oxi-
* Daniel E. Hurtado meters can be late in the detection of respiratory depression,
dhurtado@ing.puc.cl as in such condition they measure the decrease of peripheral
1
Department of Structural and Geotechnical Engineering, bloodstream oxygenation, which is known to be a late symp-
School of Engineering, and Institute for Biological tom of respiratory compromise [21]. Advanced signal pro-
and Medical Engineering, Schools of Engineering, Medicine cessing from photopletismography has been recently develop
and Biological Sciences, Pontificia Universidad Católica de to deliver a continuous estimate of RR [2]. A more accurate
Chile, Vicuña Mackenna 4860, Santiago, Chile
non-invasive approach to respiratory monitoring are systems
2
Department of Electrical Engineering, School based on measuring changes of thoracic electrical imped-
of Engineering, Pontificia Universidad Católica de Chile,
Vicuña Mackenna, 4860 Santiago, Chile ance. Some impedance-based systems have the advantage of
3 delivering not only RR but also other important ventilatory
DICTUC S.A., Vicuña Mackenna, 4860 Santiago, Chile

13
Vol.:(0123456789)
 Journal of Clinical Monitoring and Computing

parameters such as minute ventilation [3, 19]. Impedance-

based respiratory monitors, however, have the disadvantage
of being inaccurate in detecting RR in the low-frequency
range, as well as being subject to motion artifacts [6]. RR
continuous estimates based on the motion of the chest and
abdomen have also been estimated using piezoelectric sens-
ing devices [13].
An alternative method to monitor respiratory activity
is the use acoustic transducers attached to the throat of
patients. Acoustic signals acquired are then filtered, from
which RR can be determined with high accuracy [8, 17].
However, noise coming from the environment or from the
patient when he or she coughs, sneezes, snores or cries can Fig. 1  Isometric view of the respiratory sensor: (1) Case, (2) Flexible
support, (3) Nasal heat concentrator, (4) Oral heat concentrator
decrease the performance of respiratory monitors based on
acoustic sensors [7]. More recently, optical cameras and
radars have been used to contactless monitor the respiratory material with low thermal conductivity. The fixation of the
activity. Using a camera and image processing, RR has been sensor to the user skin is materialized either through an
inferred by determining the changes in color of the skin of medical-grade adhesive sticker placed in the surface that
patients [18], which has also been studied by analyzing tem- is in contact with the skin. The dimensions of the sensor
perature changes in thermal images from infrared thermal ( length × width × height)  are 30 mm × 16 mm × 20 mm
cameras [9]. RR can also be determined from chest motion and the weight is 8 g.
by tracking pixels trajectories from camera images [14] as Nasal and oral thermal collectors made of copper are
well as using radar technology [20]. While these methods installed in the housing surface facing nasal and oral flow.
provide a novel contactless alternative to standard moni- Thermal collectors transfer the heat coming from the res-
toring systems, they can be sensitive to involuntary body piration gas flow to temperature sensors, placed inside the
motions, and their estimates can be hindered in the hospital housing and in contact with the collectors. Temperature
setting when people or objects block the space between the changes of the respiratory flowing gas will result in pro-
device and the chest of the patient. portional changes in the thermal sensor resistance, which
In this paper, we present a novel continuous and non-inva- are transformed into continuous electrical signals that are
sive temperature respiratory monitoring system (TRMS) for transferred to the external monitor for further processing.
RR. Nasal and oral respiratory flow is derived from analyz- An analog-to-digital converter (ADC) integrated in a micro-
ing temperature changes occurring at the respiratory sensor, controller within the external monitor converts the analog
and then used to estimate the RR. We test the accuracy and voltage signal into digital data (10-bit resolution and 100-S/s
robustness of the system in the estimation of RR under dif- sampling rate), which is then processed to obtain the relevant
ferent guided and spontaneous respiratory settings in normal respiratory flow parameters.
volunteers. To estimate the RR the nasal and oral signals are added,
and a mean-crossing algorithm is employed. Mean-crossing
algorithms generally results in accurate estimates both for
2 Materials and methods low and high frequencies [16]. Figure 2 shows how RR is
computed using a representative graphical example. We
2.1 Description of the monitoring system use a time window of 30 s to compute the mean respiratory
cycles periods, in order to estimate the mean respiratory rate.
The temperature-based respiratory monitoring system After signal processing, the respiratory signal is graphically
consists of a respiratory sensor, and an external monitor, displayed in the screen of the external monitor, along with
where respiratory signal, respiratory frequency and other the RR.
flow parameters are displayed. The respiratory sensor is
a wearable, lightweight, non-invasive device that adheres 2.2 Position sensitivity and human breathing tests
to the skin below the nose and above the upper lip of the
subject. It measures the thermal component of the user’s Twenty healthy subjects, ten females and ten males, were
respiration gas flow and transforms temperature into an recruited for this study and underwent a protocol approved
electrical signal, which is processed and sent to the exter- by the institutional ethics committee of the Pontificia Uni-
nal monitor. An isometric view of the sensor is included versidad Católica de Chile. Inclusion criteria were: volun-
in Fig. 1. The sensor includes one housing made of plastic teers with age between 18 and 65 y old, non-smokers, no

13
Journal of Clinical Monitoring and Computing

compute the RR in a continuous way, and signal-to-noise

ratio (SNR). The protocol was repeated on the same subject
using a thermistor instead of the TRMS (NTC 10k thermis-
tor, Vishay Components, Malvern, Pennsylvania, USA), for
which the same outputs were recorded, in order to compare
both performances.
Several human breathing tests were performed on all vol-
unteers as part of the experimental protocol. First, subjects
were asked to breath through their nose for one minute fol-
lowing a metronome at rates 6, 8 and 10 BPM. This experi-
ment was then repeated for oral breathing, with nasal flow
blocked by using nose pins. Second, subjects were asked to
rest and breath spontaneously during 10 min. Third, after
a period of 1 min of spontaneous breathing, subjects were
asked to hold their breath as long as it was comfortable for
them. Finally, after a second period of one min of normal
breathing, subjects were asked to breath against a closed
Fig. 2  RR estimation using a mean-crossing algorithm glottis for as long a they could, to simulate obstructed
breathing situations.

2.3 Statistical analysis

The RR signal delivered by the monitoring system was com-

pared to the gold standard. The relative error for the moving
time window i was calculated as

RRi,TRMS − RRi,ref
Errori [%] = 1 (1)
2
(RRi,TRMS + RRi,ref )

where RRi,ref is the reference value of RR obtained from

Fig. 3  Grid and position markers for the sensitivity test
visual counting, and RRi,TRMS is the RR obtained from the
TRMS. The bias is defined as the average of the set of rela-
tive errors ( i = 1, ..., Nwindows ), the precision is the standard
record of chronic pulmonary disease, and no allergies to
deviation of the set of relative errors, and the accuracy is
copper. The breathing sensor was installed in all subjects,
defined as the root-mean-square value of the relative error
and the respiratory signal was continuously measured during
set.
the whole duration of the study. Additionally, a pulse oxi-
meter (Particle sensor, MAX30105, Sparkfun, Niwot, Colo-
rado, USA) was installed on all the subjects to continuously
record the peripheral oxygen saturation (SpO2). During the 3 Results
experiment, visual counting of respiratory cycles was elec-
tronically recorded by means of a push-button counter con- The SNR was computed for all locations in the grid consid-
trolled by an expert, which was considered the gold standard ered in the position sensitivity study. Continuous maps were
for the purpose of method comparison and error analysis. generated, by means of interpolation, for the cases of guided
All subjects were requested to lay down on a stretcher in RR = 6 BPM (Fig. 4a) and guided RR = 15 BPM (Fig. 4b),
supine position. both for the thermistor and the TRMS. A solid line is plot-
To test the sensitivity of the sensor to the position, we ted in all figures for the level set where SNR = 0 , indicat-
acquired respiratory signals at several positions in a grid ing regions where the signal and the noise have the same
marked on the face of one volunteer, as shown in Fig. 3. The amplitude. In both cases, the SNR distribution in the case
grid spacing was 7.5 mm both in the horizontal and vertical of the TRMS was positive for practically the entire spatial
directions. The subject breathing was guided by a metro- domain considered in the study. In contrast, the thermistor
nome set at 6 and 15 BPM for a period of 1 min. Respiratory had positive SNR values close to the nostrils, and quickly
flow signals acquired by the sensor were then processed to decayed as it was located farther away.

13
 Journal of Clinical Monitoring and Computing

Fig. 4  Signal-to-noise ratio map during sensitivity test. a Guided breathing at 6 BPM. b Guided breathing at 15 BPM. Solid black line shows the
level set that corresponds to SNR = 0 dB

Accuracy of the RR estimated by the TRMS and by a The respiratory signals for one subject under guided and
thermistor was computed for all locations in the grid con- spontaneous breathing are reported in Fig. 6. Bias, preci-
sidered in the position sensitivity study. Continuous maps sion and accuracy values for the sample group under guided
were generated by means of interpolation for the cases of breathing at 6, 8 and 10 BPM are reported in Table 1 for
guided RR = 6 BPM (Fig. 5a) and guided RR = 15 BPM nasal breathing and oral breathing. The mean error for nasal
(Fig. 5b), both for the thermistor and the TRMS. In the breathing and oral breathing was 0.1 and − 1.6 BPM respec-
case of the TRMS, the accuracy was less than 5% for most tively (Fig. 7). In these controlled experiments, the absolute
of the area under study, with the exception of a very small bias (average error) was less than 0.2 BPM and 2.0 BPM
region in the bottom right corner. In contrast, the ther- for nasal and oral breathing, respectively. In general, error
mistor only achieves accuracy values less than 5% in a measures for the case of oral breathing are larger than in the
reduced region directly under the nostril, out of which case of nasal breathing.
accuracy rapidly increases up to values of 40%. The time evolution of the RR for a representative subject
(subject 10) under spontaneous breathing, measured both

Fig. 5  Accuracy map during sensitivity test. a Guided breathing at 6 BPM. b Guided breathing at 15 BPM

13
Journal of Clinical Monitoring and Computing

Fig. 6  Respiratory patterns of a volunteer. a Guided breathing at 6 BPM. b Guided breathing at 8 BPM. c Guided breathing at 10 BPM. d Spon-
taneous breathing

Table 1  Error analysis for guided breathing tests. Absolute error is expressed in BPM, relative error reported in parenthesis
RR [BPM] Nasal Oral
6 8 10 6 8 10

Bias 0.1 (1.8%) 0.1 (1.5%) 0.2 (1.5%) − 0.9 (− 27.4%) − 1.9(− 39.4%) − 2.0(− 34.8%)
Bias 95% CI [0.06, 0.16] [0.06, 0.18] [0.11, 0.20] [− 1.86, 0.06] [− 3.11, − 0.75] [− 3.50, − 0.49]
Precision 0.2 (3.4%) 0.3 (3.3%) 0.3 (2.8%) 2.3 (61.0%) 2.8 (60.2%) 3.6 (62.7%)
Accuracy 0.2 (3.8%) 0.3 (3.6%) 0.3 (3.2%) 2.5 (66.9%) 3.4 (72.0%) 4.1 (71.7%)

CI confidence interval

by the TRMS and by visual counting, is reported in Fig. 8a. is − 0.04 BPM, the standard deviation is 0.4 BPM, and the
The TRMS closely follows the reference values during the 95% confidence interval is [− 0.8, 0.7] BPM.
entire time frame of the spontaneous breathing experiment. Bland-Altman analysis was also performed for the
To assess the agreement between the TRMS and the visual- whole sample of volunteers ( N = 20 ) under spontaneous
count reference, a Bland-Altman plot for the representative breathing, whose results are included in Fig. 9. In this
subject is included in Fig. 8b, where the mean RR difference case, the mean and standard deviation of the RR difference

13
 Journal of Clinical Monitoring and Computing

Fig. 7  Bland Altman plots for guided respiration tests. a Nasal breathing. b Oral breathing

Fig. 8  Results for the spontaneous breathing test for a representative subject. a Respiratory rate evolution in time for the TRMS and the refer-
ence method (visual counting). b Bland-Altman plot for assessing the agreement of RR [BPM] measured by TRMS and the reference method

between the TRMS and the reference method for the group flow was held, whereas the SpO2 was above 98% during the
was 0.4 BPM and 0.45 BPM, respectively. The 95% con- entire time frame analyzed.
fidence interval for the average RR difference was [− 0.5, During the spontaneous breathing study of subject 6,
1.3] BPM. apnea events not directed by the examiner were captured
The time evolution of the TRMS signal and RR, and the both by the TRMS system and by visual counting when
peripheral oxygen saturation (SpO2) measured by pulse oxi- the patient fell asleep. The TRMS signal and RR, and the
metry in subject 7 for the erratic breathing test is reported in peripheral oxygen saturation (SpO2) measured by pulse oxi-
Fig. 10a. From the respiratory signal, a plateau is found in metry for subject 6 are reported in Fig. 10b. Three apnea
the time interval [60,90], which corresponds to a breath hold events were recognized from the respiratory signal, where
of roughly 30 s. During that time frame, the RR measured plateaus are clearly identified. The RR measured by the
by the TRMS reached 5 BPM only 15 s after respiratory TRMS fell down to 6 BPM during the three apnea events

13
Journal of Clinical Monitoring and Computing

confirmed that the TRMS sensor provides a positive SNR for

a larger region of placement than that delivered by standard
thermistor, significantly outperforming standard thermistors
when subject to misplacement. This enhanced performance
of the TRMS is also reflected by accuracy values that were
considerably smaller than those delivered by the thermistor
system. These findings suggest that the TRMS can robustly
measure nasal respiratory signal and RR more accurately
than standard thermistors for a variety of positions, which
highlights the robustness of the TRMS to accidental mis-
placement during installation.
Spontaneous-breathing tests showed that the TRMS
provides an accurate estimate of the RR when com-
pared to visual-based RR estimates. In the sample
group, the mean RR difference obtained from the TRMS
was 0.4 ± 0.45 BPM (equivalently, bias = 0.4 BPM  ,
Fig. 9  Subject group Bland-Altman plot for assessing the agreement
precision = 0.45 BPM , and accuracy = 0.36 BPM ), which
of RR [BPM] measured by the TRMS and the reference method
is comparable with studies using an electrical-impedance
monitor on healthy subjects [19], a piezo-electric based
identified. The SpO2 was above 95% during the whole dura- monitor on PACU patients [13], and a photo-pletismogra-
tion of the study, but some oscillations were observed, which phy-based monitor on healthy subjects [1] reported in the
seemed to be temporally correlated to the TMRS RR signal. literature, where the mean RR difference were 0.1 ± 0.17
BPM, 0.41 ± 1.79 BPM and 0.23 ± 1.14 BPM, respectively.
While the mean RR difference obtained with the TRMS is
4 Discussion at the upper end of these values, we note that the standard
deviation is at the lower end, supporting the good preci-
In this work we have presented a novel temperature-based sion in estimating RR provided by the TRMS. Further,
respiratory monitoring system. A novel concept in the TRMS we note that a small standard deviation directly impacts
is the thermal collector included in the respiratory sensor, the 95% confidence interval for the mean RR difference,
made of a thermally-conductive material facing the respira- which in our case was [− 0.8,0.7] BPM. The length of this
tory gas entering and leaving the nose and mouth. From confidence interval is significantly below 3 BPM, which
our study on the position sensitivity for nasal breathing, we has been used as a measure to classify RR estimates as

Fig. 10  Results for erratic breathing conditions. a Time evolution of Time evolution of TRMS signal amplitude, respiratory frequency and
TRMS signal amplitude, respiratory frequency and pulse-oximeter pulse-oximeter SpO2 for subject 6 during a spontaneous apnea event
SpO2 for subject 7 under simulated apnea (controlled breath hold). b

13
 Journal of Clinical Monitoring and Computing

clinically relevant [13]. RR estimates obtained by the References

TRMS in spontaneous breathing subjects are, in general,
significantly more accurate than those based on visual 1. Addison PS, Watson JN, Mestek ML, Mecca RS. Developing an
imaging ( bias = 4.72 BPM [18]), near infra-red imaging algorithm for pulse oximetry derived respiratory rate (RRoxi): a
healthy volunteer study. J Clin Monit Comput. 2012;26(1):45–51.
( accuracy = 1.12 BPM [14]), and infrared thermography 2. Addison PS, Watson JN, Mestek ML, Ochs JP, Uribe Alberto A,
( bias = 1.75 BPM [9]). Bergese Sergio D. Pulse oximetry-derived respiratory rate in gen-
From the erratic breathing tests we conclude that the eral care floor patients. J Clin Monit Comput. 2014;29(1):113–20.
TRMS is capable of detecting when respiratory flow stops 3. Cavalcante AN, Martin YN, Sprung J, Imsirovic J, Weingarten
TN. Low minute ventilation episodes during anesthesia recov-
by using a threshold on the RR signal (RR < 6 BPM). Such ery following intraperitoneal surgery as detected by a non-
approach was able to capture breath hold conditions in invasive respiratory volume monitor. J Clin Monit Comput.
healthy subjects with a latency no larger than 20 s. In all 2017;32(5):929–35.
subjects, SpO2 during such trained apnea events was very 4. Cretikos MA, Bellomo R, Hillman K, Chen J, Finfer Simon, Fla-
bouris Arthas. Respiratory rate: the neglected vital sign. Med J
steady, and never reached saturation values below 95%. Aust. 2008;188(11):657–9.
An interesting observation is the case of one patient who 5. Dahan A, Aarts L, Smith TW. Incidence, reversal, and preven-
developed spontaneous apnea events while asleep. In that tion of opioid-induced respiratory depression. Anesthesiology.
case, the TRMS was able to accurately detect such events, 2010;112(1):226–38.
6. Folke M, Cernerud L, Ekström M, Hök B. Critical review of non-
which suggest that the TRMS has a potential in alerting invasive respiratory monitoring in medical care. Med Biol Eng
apnea events. Further studies that focus on sleep apnea and Comput. 2003;41(4):377–83.
respiratory depression should be carried out to evaluate in 7. Güder F, Ainla A, Redston J, Mosadegh B, Glavan Ana, Martin
depth this clinical application for the TRMS. TJ, Whitesides George M. Paper-based electrical respiration sen-
sor. Angewandte Chemie Int Ed. 2016;55(19):5727–32.
An important limitation of this study was the lower 8. Guechi Y, Pichot A, Frasca D, Rayeh-Pelardy F, Lardeur Jean-
performance of the oral sensor in estimating RR, as evi- Yves, Mimoz Olivier. Assessment of noninvasive acoustic res-
denced by the controlled-breathing experiments where piration rate monitoring in patients admitted to an emergency
the maximum error bias was − 2.0 BPM, which is large department for drug or alcoholic poisoning. J Clin Monit Comput.
2015;29(6):721–6.
when compared to the maximum bias of 0.2 BPM obtained 9. Hochhausen N, Barbosa Pereira C, Leonhardt S, Rossaint R,
for the nasal breathing case. Future developments of the Czaplik Michael. Estimating respiratory rate in post-anesthesia
TRMS should improve the temperature-based signal in care unit patients using infrared thermography: an observational
order to decrease the measurement error. This finding not- study. Sensors. 2018;18(5):1618.
10. Hogan J. Why don’t nurses monitor the respiratory rates of
withstanding, we note that the lower performance of the patients? Br J Nurs. 2006;15(9):489–92.
oral sensor does not adversely affect the TRMS estimate 11. Knaus WA, Zimmerman JE, Wagner DP, Draper EA, Lawrence
of the RR for spontaneous breathing, which can be nasal DE. APACHE-acute physiology and chronic health evaluation:
or oral. Future studies should also focus on validating the a physiologically based classification system. Crit Care Med.
1981;9(8):591–7.
performance of the TRMS in patients in a clinical setting, 12. Lee LA, Caplan RA, Stephens LS, Posner KL, Terman Greg-
such as the PACU, in order to confirm the applicability of ory  W, Voepel-Lewis Terri, Domino Karen  B. Postopera-
our system to clinical respiratory monitoring purposes. tive opioid-induced respiratory depression. Anesthesiology.
2015;122:659–65.
Acknowledgements  Authors gratefully acknowledge the support and 13. Lee PJ. Clinical evaluation of a novel respiratory rate monitor. J
assistance of Dr. Fernando Altermatt and Dr. Maria Rodriguez-Fer- Clin Monit Comput. 2016;30:175–83.
nandez in the human pilot study, and the insightful discussions about 14. Li MH, Yadollahi A, Taati B. Noncontact vision-based cardiopul-
respiratory monitoring with Dr. Carolina Cabrera. monary monitoring in different sleeping positions. IEEE J Biomed
Health Inform. 2017;21(5):1367–75.
15. Loughlin PC, Sebat F, Kellett JG. Respiratory rate: the forgot-
Funding  This research project was funded by Fundación Copec- ten vital sign - make it count!. Jt Comm J Qual Patient Saf.
UC Grant 2015.R.557 and by CORFO-Innova Grant 17ITE2-72695 2018;44(8):494–9.
awarded to DH and AA. 16. Lu W, Nystrom MM, Parikh PJ, Fooshee DR, Hubenschmidt
James P, Bradley Jeffrey D, Low Daniel A. A semi-automatic
Compliance with ethical standards  method for peak and valley detection in free-breathing respiratory
waveforms. Med Phys. 2006;33(10):3634–6.
Conflict of interest  Authors declare no conflict of interest. 17. Mimoz O, Benard T, Gaucher A, Frasca D, Debaene B. Accu-
racy of respiratory rate monitoring using a non-invasive
Ethical approval  All procedures performed in studies involving human acoustic method after general anaesthesia. Br J Anaesth.
participants were in accordance with the ethical standards of the institu- 2012;108(5):872–5.
tional ethics committee of the Pontificia Universidad Católica de Chile 18. van Gastel M, Stuijk S, de Haan G. Robust respiration detec-
(Project ID 170523007, approved October 19th, 2017). All subjects tion from remote photoplethysmography. Biomed Opt Express.
provided written informed consent to participate. 2016;7(12):4941–57.
19. Voscopoulos C, Brayanov J, Ladd D, Lalli M, Panasyuk Alexan-
der, Freeman Jenny. Evaluation of a novel noninvasive respiration

13
Journal of Clinical Monitoring and Computing

monitor providing continuous measurement of minute ventilation Review of Non-invasive Monitoring of Respiratory Condition for
in ambulatory subjects in a variety of clinical scenarios. Anesth Extubated Patients with or at Risk for Obstructive Sleep Apnea
Analg. 2013;117(1):91–100. after Surgery. Frontiers in Medicine. 2017;4(March):1–6.
20. Wang S, Pohl A, Jaeschke T, Czaplik M, Köny M, Leonhardt
S, Pohl N. A novel ultra-wideband 80 ghz fmcw radar system Publisher’s Note Springer Nature remains neutral with regard to
for contactless monitoring of vital signs. In: 2015 37th annual jurisdictional claims in published maps and institutional affiliations.
international conference of the IEEE engineering in medicine and
biology society (EMBC). IEEE; 2015. pp. 4978–4981.
2 1. Zhang Xuezheng, Kassem Mahmoud Attia Mohamed, Zhou Ying,
Shabsigh Muhammad, Wang Quanguang, Xuzhong Xu. A Brief

13