Australas Phys Eng Sci Med DOI 10.1007/s13246-014-0271-z
SCIENTIFIC PAPER
A novel device based on smart textile to control heart’s activity during exercise Marco Romagnoli • Rafael Alis • Javier Guillen • Javier Basterra • J. P. Villacastin • Sergio Guillen
Received: 18 June 2013 / Accepted: 8 April 2014 Ó Australasian College of Physical Scientists and Engineers in Medicine 2014
Abstract In recent years, several systems have been developed to control cardiac function during exercise, and some are also capable of recording RR data to provide heart rate variability (HRV) analyses. In this study we compare time between heart beats and HRV parameters obtained with a smart textile system (GOW; Weartech sl., Spain) and an electrocardiogram machine commonly used in hospitals during continuous cycling tests. Twelve cardiology patients performed a 30-min cycling test at stable submaximal intensity. RR interval data were recorded during the test by both systems. 3-min RR segments were taken to compare the time intervals between beats and HRV variables using Bland–Altman analyses and intraclass correlation coefficients. Limits of agreement (LoAs) on RR intervals were stable at around 3 ms (widest LoAs -5.754 to 6.094 ms, tightest LoAs -2.557 to 3.105 ms, medium LoAs -3.638 ± 0.812 to 3.145 ± 0.539 ms). HRV
M. Romagnoli R. Alis (&) J. Basterra Universitary Research Institute ‘‘Dr. Vin˜a Giner’’, Molecular and Mitochondrial Medicine, Catholic University of Valencia ‘‘San Vicente Ma´rtir’’, c/Quevedo 2, 46001 Valencia, Spain e-mail: [email protected] M. Romagnoli Department of Physical Education and Sports, Catholic University of Valencia ‘‘San Vicente Ma´rtir’’, Valencia, Spain J. Guillen Weartech sl., Valencia, Spain J. P. Villacastin San Carlos Clinic Hospital, Madrid, Spain S. Guillen ITACA-TSB, Valencia, Spain
parameters related to short-term change presented wide LoAs (RMSSD -0.17 to 18.41 %, HF -17.64 to 33.21 %, SD1 -0.50 to 17.54 %) as an effect of the error measurement of the GOW system. The GOW system is a valid tool for controlling HR during physical activity, although its use as a clinical tool for HRV cannot be supported. Keywords RR intervals Bland–Altman plot Heart rate variability Smart textile system
Introduction Heart rate variability (HRV) is the oscillation of the interval between consecutive heartbeats [1] and is a useful tool for providing information about regulation of the cardiac function by the autonomic nervous system [2]. It is influenced by various physiological [3] and psychological [4, 5] disorders. Its usefulness is recognized in many areas; to identify cardiac pathologies [6]; as a predictor of brain death [7, 8]; to perform trauma triage [9]. In the sports field, HRV is used to determine an athlete’s fitness level, overtraining, heart dynamics on efforts recovery [10, 11], exercise prescription [12, 13] and to identify anaerobic thresholds [14–16]. Recently, new systems meeting the sampling rate and error handling requirements for the registration of HRV [1] have been developed, and allow the subject’s free movement. These systems have shown substantial agreement with commercial electrocardiogram (ECG) systems at rest [17–20] and during exercise [21, 22]. Smart textile systems are able to record, store and/or transmit body signals [23]. There are many applications for these systems, such as monitoring emergency teams or ambulatory patient monitoring [24, 25]. These systems are
123
Australas Phys Eng Sci Med
able to register HR activity but to our knowledge there is not any smart textile system capable of transmit RR data in order to analyze HRV. The GOW system (Weartech sl., Spain), discussed in this work, consists of a smart textile shirt with sensors (textrodes) that registers the heart’s electrical impulses and transmits them to an electronic module placed on the chest. This module records and transmits the data in real-time to a device such as a personal computer or a smartphone. This system combines RR interval data collection, subject’s free movement and realtime transmission, allowing a new spectrum of potential applications. The GOW system has been designed to be applied in sports training and/or monitoring cardiac patients during their rehabilitation. Therefore, in the present study we assessed the level of concordance between the data deriving from the GOW system and an ECG commonly used in hospitals during continuous cycling tests in cardiac patients. Moreover, as systems capable of RR interval recording can be used for HRV applications, we extended our analysis to the most usual HRV parameters.
Methods Subjects Twelve adult male volunteers aged between 52 and 66 years [age 60.8 ± 5.76 years, height 174.2 ± 7.1 cm, weight 65.6 ± 6.5 kg, BMI 21.6 ± 1.5, mean ± standard deviation (SD)] participated in this study. They all suffered acute myocardial infarction. According to the potential applications of the GOW system, all the participants were patients undergoing cardiac rehabilitation after cardiac interventions or treatments. Exclusion criteria included alcoholism and smoking. Some patients were exposed to drugs that could influence the autonomic balance of heart activity, because of their condition. In all cases, the participants were monitored by medical personnel while carrying out the tests. Participants were instructed to not drink alcohol or stimulating beverages at least 12 h before the test. They were informed of the procedure and the study objectives, and they signed a consent document. The study was approved by the Ethics Committee of the San Carlos Clinic Hospital in Madrid. Data acquisition RR intervals were recorded simultaneously by the GOW system (GOW) and a commercial ECG brand Cardiolab II plus (ECG) (Prucka Engineering, TX, USA) at the 250 and 1,000 Hz resolutions, respectively. First, 10-mm AgCl surface electrodes were placed in the subject’s
123
chest (12-lead configuration) and plugged to the ECG. Then, the subjects got the GOW shirt and the GOW module was attached to the shirt. This shirt has embedded two textile electrodes in both sides of the chest that transmit the electrical activity of the heart to the module which record and/or store the signal. In this study the RR interval data were stored in the GOW module. Once the test had finished, the data collected in the module were downloaded to a personal computer. The data recorded in the ECG were down-sampled to 250 Hz and then RR interval data were exported to a personal computer. The RR data from both devices were exported to Excel 2007 (Microsoft, Inc., USA) for visual comparison and synchronization purposes. Experimental design Each subject performed the test once, under constant monitoring. The room where the test took place was conditioned at 22–24 °C and 50–60 % humidity. The test began with a 1-min warm-up of 45 W pedaling, followed by a gradual increase of load until each patient’s target HR was reached. During the test the cadence was set free between 60 and 90 rpm. The target HR was established by the medical staff in the Cardiology Department based on the patient’s pathology and physical condition, it ranged between 75 and 95 bpm. The test ended after 30 min of exercise. To limit synchronization problems between the systems, the GOW was activated first and the ECG was activated 1 min later. Data analysis Three-minute segments were selected from each subject’s RR records during exercise. The first warm-up minute was excluded. RR segments were represented to analyze and ensure synchronization between systems. Artifacts or ectopic beats were removed from RR segments. For both GOW and ECG, all the beats with a difference greater than 20 % between consecutive heartbeats were marked as artifacts, removed and replaced with interpolated values. We found substantial artifact problems for ECG recordings due to movements of the electrodes up on the subject. A total of 44 segments were analyzed (7,809 beats). For each of 3-min segment, an HRV analysis was performed with the Kubios software (University of Kuopio, Finland) [26, 27]. The smoothness priors method was used to eliminate non-stationary trends [28]. For the time domain HRV, we calculated the SD of all the normal intervals (SDNNs), the square root of the differences of successive NN intervals (RMSSD) and the mean RR intervals (MeanRR).
Australas Phys Eng Sci Med Table 1 Tightest and widest LoA for the RR intervals between ECG and GOW methods MPM (ms)
Absolute Bias
Relative LoAs
Bias (%)
LoAs (%)
Tightest LoA
812.90 ± 58.57
0.274
-2.557 to 3.105
0.03
-0.31 to 0.38
Widest LoA
721.35 ± 25.45
0.345
-5.759 to 6.094
0.05
-0.80 to 0.84
The tightest LoA correspond to the segment with lower LoA and the widest LoA is the one with higher LoA. Relative as absolute/MPM Bias absolute difference between ECG and the GOW system, MPM mean of paired means (mean ± SD)
For the frequency domain, we used Fast Fourier Transform to obtain the power spectrum density (PSD) and then calculate the power accumulated on the recommended frequency bands [1], low frequency (LF 0.04–0.15 Hz), high frequency (HF 0.15–0.40 Hz) and in total power (TP). LF/HF ratio has been reported as well as the and HF power in normalized units (HFnu) calculated as (HF/ (HF ? LF)) 9 100. Prior to calculating the PSD, the interval series were interpolated with the Cubic Spline method, set at 4 Hz to transform them into equidistant samples. The Poincare´ plot displays the correlation between consecutive intervals and typically appears as an elongated cloud of points oriented along the line-of-identity [29]. This analysis method was included due to its applicability to non-stationary data sets [26]. An ellipse can be fitted to the Poincare´ plot, being SD1 the standard deviation of the distances from points to the mayor axis of the ellipse and SD2 the standard deviation of the distances from points to the minor axis of the ellipse [30]. SD1 is a measure of short-term HRV and SD2 measures long-term HRV [29]. Therefore, we considered SD1 and SD2 coefficients of the Poincare´ plot [29, 31] of each segment. Statistical analysis Data were analyzed for normality using the Kolmogorov– Smirnov test. To analyze the agreement between methods on the RR recording for each segment and the HRV analysis data, we followed the Bland–Altman method [32] calculating the limits of agreement (LoAs) at 95 % based on the mean difference. The bias and LoAs were expressed as absolute values and as relative splitting the absolute values by the mean of paired means of the values for each method. A priori acceptable relative LoAs were considered below 10 %. Additionally, to analyze the concordance between methods we calculated the intraclass correlation coefficients (ICCs) with 95 % confidence intervals (95 % CIs). A good concordance between methods was considered when the low 95 % CI of ICC was higher than 0.8 [33]. The statistical analysis was performed using the SPSS version 19 (SPSS, Inc., Chicago, IL, USA). Statistical significance was set at p \ 0.05. Data are presented as mean and standard deviation (mean ± SD).
Fig. 1 Bland–Altman plots of the segments with tightest (upper) and widest (lower) LoAs. The tightest LoA correspond to the segment with lower LoA and the widest LoA is the one with higher LoA
Results There were no-wide discrepancies in the LoAs of the RR intervals recorded by the two systems (Table 1; Fig. 1). After considering all the segments, the LoAs were around ±3 ms (LoA inf -3.638 ± 0.812 ms, LoA sup 3.145 ± 0.539 ms). Table 2 contains the LoAs for the HRV parameters calculated for the both systems. MeanRR, SDNN and SD2 present excellent LoAs for the absolute
123
Australas Phys Eng Sci Med Table 2 LoAs for the HRV parameters between ECG and GOW methods MPM
Absolute Bias
MeanRR (ms)
Relative LoAs
Bias (%)
LoAs (%)
703.75 ± 49.67
0.20
0.02 to 0.38
0.02
0.002 to 0.05
SDNN (ms)
1.14 ± 0.56
0.02
-0.01 to 0.05
1.75
-0.87 to 3.51
RMSSD
5.81 ± 2.01
0.53
-0.01 to 1.07
9.12
-0.17 to 18.41
TP (ms )
39.40 ± 31.91
0.71
-2.14 to 3.56
1.77
-5.35 to 8.91
LF (ms2)
25.42 ± 23.65
0.06
-1.42 to 1.54
0.23
-5.58 to 6.05
HF (ms2)
5.78 ± 6.12
0.45
-1.02 to 1.92
7.78
LF/HF HFnu (%)
8.57 ± 8.09 19.95 ± 18.42
2
-2.51 3.31
-9.03 to 4.01 -5.41 to 12.03
-29.35 16.60
-17.64 to 33.21 -105.40 to 46.73 -27.10 to 60.30
SD1 (ms)
3.99 ± 1.95
0.34
-0.02 to 0.70
8.52
-0.50 to 17.54
SD2 (ms)
24.01 ± 15.88
0.02
-0.26 to 0.30
0.08
-1.08 to 1.24
Relative as absolute/MPM Bias absolute difference between ECG and the GOW system, MPM mean of paired means (mean ± SD)
and relative values, but RMSSD, HF, LF/HF, HFnu and SD1 showed a worse agreement between systems (Table 2; Fig. 2). Table 3 contains the ICCs values. For RMSSD and SD1, the low 95 % CI of ICC were less than 0.8 (Table 3).
Discussion This study is the first to compare the time intervals between the heartbeats and HRV parameters obtained from a commercial ECG and from a smart textile system. Other authors have performed comparative studies of alternative measurement systems such as the Polar [17–19, 22, 34] or Suunto [21] HR monitor. The results of the present study show good agreement regarding RR intervals and unequal results for the HRV parameters during low intensity cycling. The LoAs for the RR intervals obtained by the Bland– Altman method were at about ±3 ms (LoA inf 3.638 ± 0.812 ms, LoA sup 3.145 ± 0.539 ms). The narrower and wider LoAs for all the segments (Table 1; Fig. 1) were similar to those obtained in a study with healthy subjects with a Polar s810i in a cycling incremental test [22]. In this study, the LoAs increases with the exercise intensity increasing. However, in the present study, no trend increasing LoAs while increasing HR during the test has been found. The MeanRR variable presents low absolute LoAs (between 0.02 and 0.38 ms). These results, along with the LoAs obtained for RR intervals shows that the GOW system provides an excellent control of HR during exercise. SDNN presented LoAs between -0.01 and 0.04 ms on absolute values and between -0.87 and 3.51 % on relative values. SDNN represents all the HRV components in the period recorded [1]. After taking into account the global
123
nature of the variable and the low LoAs values, we considered them acceptable. RMSSD estimates short-term HRV [1]. The LoAs for this variable cannot be considered acceptable (-0.17 to 18.41 % on relative values). MeanRR and SDNN values showed an excellent correlation (Table 3) and therefore concordance between methods. RMSSD showed a high correlation (ICC = 0.960), but the low 95 % CI of ICC was less than 0.80 (Table 3). Therefore, the two methods do not show good concordance to measure this variable. The agreement between methods for this variable was bad, although the ICC was high. In this study, we decided to consider correlation and agreement for the comparison made between the two methods [35]. Therefore, we performed two statistical tests to ensure a proper comparison between systems. The Bland–Altman test provides a measure of agreement irrespectively of the probability values, whereas the ICC provides a measure of correlation of the results obtained by both systems. We consider agreement values to be more representatives to compare the systems. Although RMSSD showed a high correlation (Table 3) the LoAs are too broad (Table 2). Thus, the comparison is not favorable. The LoAs obtained for the LF parameter in the present study were similar to those obtained by Kingsley et al. [22] in absolute values (-2.7 to 2.3 vs. -1.42 to 1.54 ms2) at a comparable exercise intensity, but significantly lower ones in relative values (-2.1 to 1.8 vs. -5.58 to 6.05 %). The LF LoAs seems to be too broad. However, relative LoAs have to be carefully considered as they are calculated as absolute values divided by the mean of paired means. Especially, in the case of the mentioned parameter that shows high SDs in both methods (ECG LF 25.36 ± 22.41 ms2, GOW LF 26.00 ± 24.17 ms2). Moreover, the Bland–Altman plot (Fig. 2) shows that both methods have good agreement except in the case of two segments where the GOW system
Australas Phys Eng Sci Med Fig. 2 Bland–Altman plots of the HRV parameters
123
Australas Phys Eng Sci Med Table 3 Intraclass correlation coefficients for the HRV variables of both methods
MeanRR (ms) SDNN (ms)
ECG
GOW
ICC (95 % CI)
703.89 ± 50.95
703.37 ± 49.09
1.000 (0.998–1.000)
1.13 ± 0.80
1.15 ± 0.37
0.998 (0.988–1.000)
RMSSD
5.35 ± 2.05
6.01 ± 2.23
0.960 (0.389–0.990)
TP (ms2)
39.64 ± 31.61
40.35 ± 32.21
0.999 (0.997–1.000)
LF (ms2)
25.36 ± 22.41
26.00 ± 24.17
1.000 (0.999–1.000)
2
HF (ms )
5.59 ± 6.15
6.23 ± 6.46
0.990 (0.978–0.998)
LF/HF
9.83 ± 8.54
7.31 ± 7.96
0.919 (0.815–0.965)
Data represented as mean ± SD
HFnu (%)
18.29 ± 18.85
21.60 ± 18.22
0.971 (0.932–0.988)
ICC intraclass correlation coefficient, 95 % CI 95 % confidence interval
SD1 (ms)
3.87 ± 1.98
4.32 ± 1.38
0.948 (0.254–0.983)
SD2 (ms)
23.98 ± 14.80
24.32 ± 15.90
0.999 (0.992–1.000)
under estimate the LF parameter. As for the LF variable, the agreement between methods for the TP variable can be considered as acceptable (Table 2; Fig. 2). The LoAs for HF, LF/HF and HFnu were very large in both absolute and relative terms (Table 2; Fig. 2). All the power-spectrum related HRV parameters presented good ICCs (Table 3). However, in relation to the LoAs discussed above, we cannot conclude that the comparisons made for HF, LF/HF and HFnu parameters are favorable. SD1 presented wide LoAs for absolute and relative values (-0.02 to 0.70 ms, -0.50 to 17.54 %, respectively). In the case of SD2, the LoAs for absolute and relative values were tighter (-0.26 to 0.30 ms, -1.08 to 1.24 %, respectively). SD1 refers to the short-term variability [30] and is associated with the RMSSD and HF parameters, while SD2 is a measure of slow HRV [30]. In a study performed at rest [19], higher LoAs for SD1 than for SD2 were also found, in both the supine and standing positions. SD2 show an excellent low 95 % CI of ICC (0.992), although for SD1 was too low (0.254). These data are consistent with those observed in the concordance analysis. The results of the present study shows that the measurement error of the GOW system compared to the gold standard are broader in the case of the HRV parameters related to short time change (e.g., RMSSD, HF, and SD1). Therefore, rapid oscillations of HR showed worst LoAs whereas low oscillations seem to be more stable in the GOW system (Table 2). There are some important technical differences between both methods that can account for these differences. The sample rate of the GOW system is 250 Hz while ECG works at 1,000 Hz. Although we have down sampled the ECG data prior to RR detection, the low sample rate of the GOW system can be an important source of error measurement. Other difference between both methods is the R wave detection RR algorithm. Weippert et al. [21] have previously reported that it can be an important source of discrepancies, especially for the short-term change related HRV parameters. Moreover, the GOW system and the ECG had different electrodes
123
distribution. The GOW shirt has the textrodes disposed in both sides of the chest. None of ECG derivations correspond to this distribution. Therefore, the heart electrical activity was not recorded from the same location and this should add to the GOW error of measurement. The GOW system was designed to control training sessions of healthy subjects and to monitoring rehabilitation programs of cardiac patients. HRV analysis is particularly important in the case of cardiac patients since these data provides insight of the heart response to exercise and to the rehabilitation process [36, 37]. The results of the present study show that the measurement error of the GOW system is too large in order to accept the HRV data during exercise and therefore, do not meet the requirements as clinical tool. However, the GOW system does meet the requirements to controlling HR during exercise inasmuch as the agreement for the RR data and MeanRR variable is acceptable. In conclusion, the GOW system is a valid tool for controlling HR while performing physical activity, although its use as a clinical tool for HRV cannot be supported. Acknowledgments The authors thank Alicia Ricart for revising the level of English of this manuscript. Conflict of interest Javier Guillen was working in Weartech sl. at the time the research was conducted.
References 1. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (1996) Heart rate variability: standards of measurement, physiological interpretation, and clinical use. Eur Heart J 17(3):354–381 2. Buchheit M, Gindre C (2006) Cardiac parasympathetic regulation: respective associations with cardiorespiratory fitness and training load. Am J Physiol Heart Circ Physiol 291(1):H451– H458. doi:10.1152/ajpheart.00008.2006 3. Zhang J, Tu W, Dai J, Lv Q, Yang X (2011) Dysfunction of preand post-operative cardiac autonomic nervous system in elderly
Australas Phys Eng Sci Med
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
patients with diabetes mellitus. Arch Gerontol Geriatr 53(3): 334–337 Carney RM, Blumenthal JA, Stein PK, Watkins L, Catellier D, Berkman LF, Czajkowski SM, O’Connor C, Stone PH, Freedland KE (2001) Depression, heart rate variability, and acute myocardial infarction. Circulation 104(17):2024–2028. doi:10.1161/ hc4201.097834 Jindal RD, Vasko RC, Jennings JR, Fasiczka AL, Thase ME, Reynolds CF (2008) Heart rate variability in depressed elderly. Am J Geriatr Psychiatry 16(11):861–866. doi:10.1097/JGP. 0b013e318180053d Karp E, Shiyovich A, Zahger D, Gilutz H, Grosbard A, Katz A (2009) Ultra-short-term heart rate variability for early risk stratification following acute ST-elevation myocardial infarction. Cardiology 114(4):275–283 Goldstein B, Toweill D, Lai S, Sonnenthal K, Kimberly B (1998) Uncoupling of the autonomic and cardiovascular systems in acute brain injury. Am J Physiol Regul Integr Comp Physiol 275(4):R1287–R1292 Biswas AK, Scott WA, Sommerauer JF, Luckett PM (2000) Heart rate variability after acute traumatic brain injury in children. Crit Care Med 28(12):3907–3912 Batchinsky AI, Cancio LC, Salinas J, Kuusela T, Cooke WH, Wang JJ, Boehme M, Convertino VA, Holcomb JB (2007) Prehospital loss of R-to-R interval complexity is associated with mortality in trauma patients. J Trauma 63(3):512–518. doi:10. 1097/TA.0b013e318142d2f0 Al Haddad H, Laursen PB, Chollet D, Ahmaidi S, Buchheit M (2011) Reliability of resting and postexercise heart rate measures. Int J Sports Med 32(08):598–605. doi:10.1055/s-00311275356 Arduini A, Gomez-Cabrera M-C, Romagnoli M (2011) Reliability of different models to assess heart rate recovery after submaximal bicycle exercise. J Sci Med Sport 14(4):352–357. doi:10.1016/j.jsams.2011.02.012 Kiviniemi A, Hautala A, Kinnunen H, Tulppo M (2007) Endurance training guided individually by daily heart rate variability measurements. Eur J Appl Physiol 101(6):743–751. doi:10.1007/ s00421-007-0552-2 Kiviniemi AM, Hautala AJ, Kinnunen H, Nissila¨ J, Virtanen P, Karjalainen J, Tulppo MP (2010) Daily exercise prescription on the basis of HR variability among men and women. Med Sci Sports Exerc 42(7):1355–1363. doi:10.1249/MSS.0b013e3181 cd5f39 Cottin F, Lepreˆtre PM, Lopes P, Papelier Y, Me´digue C, Billat V (2006) Assessment of ventilatory thresholds from heart rate variability in well-trained subjects during cycling. Int J Sports Med 27(12):959–967. doi:10.1055/s-2006-923849 Cottin F, Me´digue C, Lopes P, Lepreˆtre PM, Heubert R, Billat V (2007) Ventilatory thresholds assessment from heart rate variability during an incremental exhaustive running test. Int J Sports Med 28:287–294. doi:10.1055/s-2006-924355 Blain G, Meste O, Bouchard T, Bermon S (2005) Assessment of ventilatory thresholds during graded and maximal exercise test using time varying analysis of respiratory sinus arrhythmia. Br J Sport Med 39(7):448–452. doi:10.1136/bjsm.2004. 014134 Gamelin FX, Berthoin S, Bosquet L (2006) Validity of the Polar S810 heart rate monitor to measure R–R intervals at rest. Med Sci Sports Exerc 38(5):887–893. doi:10.1249/01.mss.0000218135. 79476.9c Gamelin FX, Baquet G, Berthoin S, Bosquet L (2008) Validity of the Polar S810 to measure R–R intervals in children. Int J Sports Med 29(2):134–138. doi:10.1055/s-2007-964995
19. Nunan D, Jakovljevic D, Donovan G, Hodges L, Sandercock G, Brodie D (2008) Levels of agreement for RR intervals and shortterm heart rate variability obtained from the Polar S810 and an alternative system. Eur J Appl Physiol 103(5):529–537. doi:10. 1007/s00421-008-0742-6 20. Walle´n M, Hasson D, Theorell T, Canlon B, Osika W (2012) Possibilities and limitations of the Polar RS800 in measuring heart rate variability at rest. Eur J Appl Physiol 112(3): 1153–1165. doi:10.1007/s00421-011-2079-9 21. Weippert M, Kumar M, Kreuzfeld S, Arndt D, Rieger A, Stoll R (2010) Comparison of three mobile devices for measuring R–R intervals and heart rate variability: Polar S810i, Suunto t6 and an ambulatory ECG system. Eur J Appl Physiol 109(4):779–786. doi:10.1007/s00421-010-1415-9 22. Kingsley M, Lewis MJ, Marson RE (2005) Comparison of Polar 810s and an ambulatory ECG system for RR interval measurement during progressive exercise. Int J Sports Med 26:39–44. doi:10.1055/s-2004-817878 23. Lymberis A, Paradiso R (2008) Smart fabrics and interactive textile enabling wearable personal applications: R&D state of the art and future challenges. Conf Proc IEEE Eng Med Biol Soc 2008:5270–5273 24. Curone D, Secco EL, Tognetti A, Loriga G, Dudnik G, Risatti M, Whyte R, Bonfiglio A, Magenes G (2010) Smart garments for emergency operators: the proeTEX project. IEEE Trans Inf Technol Biomed 14(3):694–701. doi:10.1109/titb.2010.2045003 25. Villalba E, Arredondo MT, Moreno A, Salvi D, Guillen S (2006) User interaction design and development of a heart failure management system based on wearable and information technologies. Conf Proc IEEE Eng Med Biol Soc 2006:400–403 26. Grant CC, Viljoen M, van Rensburg DCJ, Wood PS (2012) Heart rate variability assessment of the effect of physical training on autonomic cardiac control. Ann Noninvasive Electrocardiol 17(3):219–229. doi:10.1111/j.1542-474X.2012.00511.x 27. Niskanen J-P, Tarvainen MP, Ranta-aho PO, Karjalainen PA (2004) Software for advanced HRV analysis. Comput Methods Program Biomed 76(1):73–81. doi:10.1016/j.cmpb.2004.03.004 28. Tarvainen MP, Ranta-aho PO, Karjalainen PA (2002) An advanced detrending method with application to HRV analysis. IEEE Trans Biomed Eng 49(2):172–175. doi:10.1109/10.979357 29. Brennan M, Palaniswami M, Kamen P (2001) Do existing measures of Poincare plot geometry reflect nonlinear features of heart rate variability? IEEE Trans Biomed Eng 48(11):1342–1347. doi:10.1109/10.959330 30. Tulppo MP, Makikallio TH, Takala TE, Seppanen T, Huikuri HV (1996) Quantitative beat-to-beat analysis of heart rate dynamics during exercise. Am J Physiol Heart C 271(1):H244–H252 31. Carrasco M, Gaita´n R, Gonza´lez O, Ya´nez S (2001) Correlation among Poincare´ plot indexes and time and frequency domain measures of heart rate variability. J Med Eng Technol 25(6):240–248. doi:10.1080/03091900110086651 32. Bland JM, Altman DG (2007) Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 17(4):571–582 33. Pinna GD, Maestri R, Torunski A, Danilowicz-Szymanowicz L, Szwoch M, La Rovere MT, Raczak G (2007) Heart rate variability measures: a fresh look at reliability. Clin Sci 113(3): 131–140. doi:10.1042/cs20070055 34. Nunan D, Donovan G, Jakovljevic DG, Hodges LD, Sandercock GRH, Brodie DA (2009) Validity and reliability of short-term heart-rate variability from the Polar S810. Med Sci Sports Exerc 41(1):243–250 35. Bland JM, Altman D (1986) Statistical Methods for assessing agreement between two methods of clinical measurement. Lancet 327(8476):307–310
123
Australas Phys Eng Sci Med 36. Routledge FS, Campbell TS, McFetridge-Durdle JA, Bacon SL (2010) Improvements in heart rate variability with exercise therapy. Can J Cardiol 26(6):303–312
123
37. Rosenwinkel ET, Bloomfield DM, Arwady MA, Goldsmith RL (2001) Exercise and autonomic function in health and cardiovascular disease. Cardiol Clin 19(3):369–387
SCIENTIFIC PAPER
A novel device based on smart textile to control heart’s activity during exercise Marco Romagnoli • Rafael Alis • Javier Guillen • Javier Basterra • J. P. Villacastin • Sergio Guillen
Received: 18 June 2013 / Accepted: 8 April 2014 Ó Australasian College of Physical Scientists and Engineers in Medicine 2014
Abstract In recent years, several systems have been developed to control cardiac function during exercise, and some are also capable of recording RR data to provide heart rate variability (HRV) analyses. In this study we compare time between heart beats and HRV parameters obtained with a smart textile system (GOW; Weartech sl., Spain) and an electrocardiogram machine commonly used in hospitals during continuous cycling tests. Twelve cardiology patients performed a 30-min cycling test at stable submaximal intensity. RR interval data were recorded during the test by both systems. 3-min RR segments were taken to compare the time intervals between beats and HRV variables using Bland–Altman analyses and intraclass correlation coefficients. Limits of agreement (LoAs) on RR intervals were stable at around 3 ms (widest LoAs -5.754 to 6.094 ms, tightest LoAs -2.557 to 3.105 ms, medium LoAs -3.638 ± 0.812 to 3.145 ± 0.539 ms). HRV
M. Romagnoli R. Alis (&) J. Basterra Universitary Research Institute ‘‘Dr. Vin˜a Giner’’, Molecular and Mitochondrial Medicine, Catholic University of Valencia ‘‘San Vicente Ma´rtir’’, c/Quevedo 2, 46001 Valencia, Spain e-mail: [email protected] M. Romagnoli Department of Physical Education and Sports, Catholic University of Valencia ‘‘San Vicente Ma´rtir’’, Valencia, Spain J. Guillen Weartech sl., Valencia, Spain J. P. Villacastin San Carlos Clinic Hospital, Madrid, Spain S. Guillen ITACA-TSB, Valencia, Spain
parameters related to short-term change presented wide LoAs (RMSSD -0.17 to 18.41 %, HF -17.64 to 33.21 %, SD1 -0.50 to 17.54 %) as an effect of the error measurement of the GOW system. The GOW system is a valid tool for controlling HR during physical activity, although its use as a clinical tool for HRV cannot be supported. Keywords RR intervals Bland–Altman plot Heart rate variability Smart textile system
Introduction Heart rate variability (HRV) is the oscillation of the interval between consecutive heartbeats [1] and is a useful tool for providing information about regulation of the cardiac function by the autonomic nervous system [2]. It is influenced by various physiological [3] and psychological [4, 5] disorders. Its usefulness is recognized in many areas; to identify cardiac pathologies [6]; as a predictor of brain death [7, 8]; to perform trauma triage [9]. In the sports field, HRV is used to determine an athlete’s fitness level, overtraining, heart dynamics on efforts recovery [10, 11], exercise prescription [12, 13] and to identify anaerobic thresholds [14–16]. Recently, new systems meeting the sampling rate and error handling requirements for the registration of HRV [1] have been developed, and allow the subject’s free movement. These systems have shown substantial agreement with commercial electrocardiogram (ECG) systems at rest [17–20] and during exercise [21, 22]. Smart textile systems are able to record, store and/or transmit body signals [23]. There are many applications for these systems, such as monitoring emergency teams or ambulatory patient monitoring [24, 25]. These systems are
123
Australas Phys Eng Sci Med
able to register HR activity but to our knowledge there is not any smart textile system capable of transmit RR data in order to analyze HRV. The GOW system (Weartech sl., Spain), discussed in this work, consists of a smart textile shirt with sensors (textrodes) that registers the heart’s electrical impulses and transmits them to an electronic module placed on the chest. This module records and transmits the data in real-time to a device such as a personal computer or a smartphone. This system combines RR interval data collection, subject’s free movement and realtime transmission, allowing a new spectrum of potential applications. The GOW system has been designed to be applied in sports training and/or monitoring cardiac patients during their rehabilitation. Therefore, in the present study we assessed the level of concordance between the data deriving from the GOW system and an ECG commonly used in hospitals during continuous cycling tests in cardiac patients. Moreover, as systems capable of RR interval recording can be used for HRV applications, we extended our analysis to the most usual HRV parameters.
Methods Subjects Twelve adult male volunteers aged between 52 and 66 years [age 60.8 ± 5.76 years, height 174.2 ± 7.1 cm, weight 65.6 ± 6.5 kg, BMI 21.6 ± 1.5, mean ± standard deviation (SD)] participated in this study. They all suffered acute myocardial infarction. According to the potential applications of the GOW system, all the participants were patients undergoing cardiac rehabilitation after cardiac interventions or treatments. Exclusion criteria included alcoholism and smoking. Some patients were exposed to drugs that could influence the autonomic balance of heart activity, because of their condition. In all cases, the participants were monitored by medical personnel while carrying out the tests. Participants were instructed to not drink alcohol or stimulating beverages at least 12 h before the test. They were informed of the procedure and the study objectives, and they signed a consent document. The study was approved by the Ethics Committee of the San Carlos Clinic Hospital in Madrid. Data acquisition RR intervals were recorded simultaneously by the GOW system (GOW) and a commercial ECG brand Cardiolab II plus (ECG) (Prucka Engineering, TX, USA) at the 250 and 1,000 Hz resolutions, respectively. First, 10-mm AgCl surface electrodes were placed in the subject’s
123
chest (12-lead configuration) and plugged to the ECG. Then, the subjects got the GOW shirt and the GOW module was attached to the shirt. This shirt has embedded two textile electrodes in both sides of the chest that transmit the electrical activity of the heart to the module which record and/or store the signal. In this study the RR interval data were stored in the GOW module. Once the test had finished, the data collected in the module were downloaded to a personal computer. The data recorded in the ECG were down-sampled to 250 Hz and then RR interval data were exported to a personal computer. The RR data from both devices were exported to Excel 2007 (Microsoft, Inc., USA) for visual comparison and synchronization purposes. Experimental design Each subject performed the test once, under constant monitoring. The room where the test took place was conditioned at 22–24 °C and 50–60 % humidity. The test began with a 1-min warm-up of 45 W pedaling, followed by a gradual increase of load until each patient’s target HR was reached. During the test the cadence was set free between 60 and 90 rpm. The target HR was established by the medical staff in the Cardiology Department based on the patient’s pathology and physical condition, it ranged between 75 and 95 bpm. The test ended after 30 min of exercise. To limit synchronization problems between the systems, the GOW was activated first and the ECG was activated 1 min later. Data analysis Three-minute segments were selected from each subject’s RR records during exercise. The first warm-up minute was excluded. RR segments were represented to analyze and ensure synchronization between systems. Artifacts or ectopic beats were removed from RR segments. For both GOW and ECG, all the beats with a difference greater than 20 % between consecutive heartbeats were marked as artifacts, removed and replaced with interpolated values. We found substantial artifact problems for ECG recordings due to movements of the electrodes up on the subject. A total of 44 segments were analyzed (7,809 beats). For each of 3-min segment, an HRV analysis was performed with the Kubios software (University of Kuopio, Finland) [26, 27]. The smoothness priors method was used to eliminate non-stationary trends [28]. For the time domain HRV, we calculated the SD of all the normal intervals (SDNNs), the square root of the differences of successive NN intervals (RMSSD) and the mean RR intervals (MeanRR).
Australas Phys Eng Sci Med Table 1 Tightest and widest LoA for the RR intervals between ECG and GOW methods MPM (ms)
Absolute Bias
Relative LoAs
Bias (%)
LoAs (%)
Tightest LoA
812.90 ± 58.57
0.274
-2.557 to 3.105
0.03
-0.31 to 0.38
Widest LoA
721.35 ± 25.45
0.345
-5.759 to 6.094
0.05
-0.80 to 0.84
The tightest LoA correspond to the segment with lower LoA and the widest LoA is the one with higher LoA. Relative as absolute/MPM Bias absolute difference between ECG and the GOW system, MPM mean of paired means (mean ± SD)
For the frequency domain, we used Fast Fourier Transform to obtain the power spectrum density (PSD) and then calculate the power accumulated on the recommended frequency bands [1], low frequency (LF 0.04–0.15 Hz), high frequency (HF 0.15–0.40 Hz) and in total power (TP). LF/HF ratio has been reported as well as the and HF power in normalized units (HFnu) calculated as (HF/ (HF ? LF)) 9 100. Prior to calculating the PSD, the interval series were interpolated with the Cubic Spline method, set at 4 Hz to transform them into equidistant samples. The Poincare´ plot displays the correlation between consecutive intervals and typically appears as an elongated cloud of points oriented along the line-of-identity [29]. This analysis method was included due to its applicability to non-stationary data sets [26]. An ellipse can be fitted to the Poincare´ plot, being SD1 the standard deviation of the distances from points to the mayor axis of the ellipse and SD2 the standard deviation of the distances from points to the minor axis of the ellipse [30]. SD1 is a measure of short-term HRV and SD2 measures long-term HRV [29]. Therefore, we considered SD1 and SD2 coefficients of the Poincare´ plot [29, 31] of each segment. Statistical analysis Data were analyzed for normality using the Kolmogorov– Smirnov test. To analyze the agreement between methods on the RR recording for each segment and the HRV analysis data, we followed the Bland–Altman method [32] calculating the limits of agreement (LoAs) at 95 % based on the mean difference. The bias and LoAs were expressed as absolute values and as relative splitting the absolute values by the mean of paired means of the values for each method. A priori acceptable relative LoAs were considered below 10 %. Additionally, to analyze the concordance between methods we calculated the intraclass correlation coefficients (ICCs) with 95 % confidence intervals (95 % CIs). A good concordance between methods was considered when the low 95 % CI of ICC was higher than 0.8 [33]. The statistical analysis was performed using the SPSS version 19 (SPSS, Inc., Chicago, IL, USA). Statistical significance was set at p \ 0.05. Data are presented as mean and standard deviation (mean ± SD).
Fig. 1 Bland–Altman plots of the segments with tightest (upper) and widest (lower) LoAs. The tightest LoA correspond to the segment with lower LoA and the widest LoA is the one with higher LoA
Results There were no-wide discrepancies in the LoAs of the RR intervals recorded by the two systems (Table 1; Fig. 1). After considering all the segments, the LoAs were around ±3 ms (LoA inf -3.638 ± 0.812 ms, LoA sup 3.145 ± 0.539 ms). Table 2 contains the LoAs for the HRV parameters calculated for the both systems. MeanRR, SDNN and SD2 present excellent LoAs for the absolute
123
Australas Phys Eng Sci Med Table 2 LoAs for the HRV parameters between ECG and GOW methods MPM
Absolute Bias
MeanRR (ms)
Relative LoAs
Bias (%)
LoAs (%)
703.75 ± 49.67
0.20
0.02 to 0.38
0.02
0.002 to 0.05
SDNN (ms)
1.14 ± 0.56
0.02
-0.01 to 0.05
1.75
-0.87 to 3.51
RMSSD
5.81 ± 2.01
0.53
-0.01 to 1.07
9.12
-0.17 to 18.41
TP (ms )
39.40 ± 31.91
0.71
-2.14 to 3.56
1.77
-5.35 to 8.91
LF (ms2)
25.42 ± 23.65
0.06
-1.42 to 1.54
0.23
-5.58 to 6.05
HF (ms2)
5.78 ± 6.12
0.45
-1.02 to 1.92
7.78
LF/HF HFnu (%)
8.57 ± 8.09 19.95 ± 18.42
2
-2.51 3.31
-9.03 to 4.01 -5.41 to 12.03
-29.35 16.60
-17.64 to 33.21 -105.40 to 46.73 -27.10 to 60.30
SD1 (ms)
3.99 ± 1.95
0.34
-0.02 to 0.70
8.52
-0.50 to 17.54
SD2 (ms)
24.01 ± 15.88
0.02
-0.26 to 0.30
0.08
-1.08 to 1.24
Relative as absolute/MPM Bias absolute difference between ECG and the GOW system, MPM mean of paired means (mean ± SD)
and relative values, but RMSSD, HF, LF/HF, HFnu and SD1 showed a worse agreement between systems (Table 2; Fig. 2). Table 3 contains the ICCs values. For RMSSD and SD1, the low 95 % CI of ICC were less than 0.8 (Table 3).
Discussion This study is the first to compare the time intervals between the heartbeats and HRV parameters obtained from a commercial ECG and from a smart textile system. Other authors have performed comparative studies of alternative measurement systems such as the Polar [17–19, 22, 34] or Suunto [21] HR monitor. The results of the present study show good agreement regarding RR intervals and unequal results for the HRV parameters during low intensity cycling. The LoAs for the RR intervals obtained by the Bland– Altman method were at about ±3 ms (LoA inf 3.638 ± 0.812 ms, LoA sup 3.145 ± 0.539 ms). The narrower and wider LoAs for all the segments (Table 1; Fig. 1) were similar to those obtained in a study with healthy subjects with a Polar s810i in a cycling incremental test [22]. In this study, the LoAs increases with the exercise intensity increasing. However, in the present study, no trend increasing LoAs while increasing HR during the test has been found. The MeanRR variable presents low absolute LoAs (between 0.02 and 0.38 ms). These results, along with the LoAs obtained for RR intervals shows that the GOW system provides an excellent control of HR during exercise. SDNN presented LoAs between -0.01 and 0.04 ms on absolute values and between -0.87 and 3.51 % on relative values. SDNN represents all the HRV components in the period recorded [1]. After taking into account the global
123
nature of the variable and the low LoAs values, we considered them acceptable. RMSSD estimates short-term HRV [1]. The LoAs for this variable cannot be considered acceptable (-0.17 to 18.41 % on relative values). MeanRR and SDNN values showed an excellent correlation (Table 3) and therefore concordance between methods. RMSSD showed a high correlation (ICC = 0.960), but the low 95 % CI of ICC was less than 0.80 (Table 3). Therefore, the two methods do not show good concordance to measure this variable. The agreement between methods for this variable was bad, although the ICC was high. In this study, we decided to consider correlation and agreement for the comparison made between the two methods [35]. Therefore, we performed two statistical tests to ensure a proper comparison between systems. The Bland–Altman test provides a measure of agreement irrespectively of the probability values, whereas the ICC provides a measure of correlation of the results obtained by both systems. We consider agreement values to be more representatives to compare the systems. Although RMSSD showed a high correlation (Table 3) the LoAs are too broad (Table 2). Thus, the comparison is not favorable. The LoAs obtained for the LF parameter in the present study were similar to those obtained by Kingsley et al. [22] in absolute values (-2.7 to 2.3 vs. -1.42 to 1.54 ms2) at a comparable exercise intensity, but significantly lower ones in relative values (-2.1 to 1.8 vs. -5.58 to 6.05 %). The LF LoAs seems to be too broad. However, relative LoAs have to be carefully considered as they are calculated as absolute values divided by the mean of paired means. Especially, in the case of the mentioned parameter that shows high SDs in both methods (ECG LF 25.36 ± 22.41 ms2, GOW LF 26.00 ± 24.17 ms2). Moreover, the Bland–Altman plot (Fig. 2) shows that both methods have good agreement except in the case of two segments where the GOW system
Australas Phys Eng Sci Med Fig. 2 Bland–Altman plots of the HRV parameters
123
Australas Phys Eng Sci Med Table 3 Intraclass correlation coefficients for the HRV variables of both methods
MeanRR (ms) SDNN (ms)
ECG
GOW
ICC (95 % CI)
703.89 ± 50.95
703.37 ± 49.09
1.000 (0.998–1.000)
1.13 ± 0.80
1.15 ± 0.37
0.998 (0.988–1.000)
RMSSD
5.35 ± 2.05
6.01 ± 2.23
0.960 (0.389–0.990)
TP (ms2)
39.64 ± 31.61
40.35 ± 32.21
0.999 (0.997–1.000)
LF (ms2)
25.36 ± 22.41
26.00 ± 24.17
1.000 (0.999–1.000)
2
HF (ms )
5.59 ± 6.15
6.23 ± 6.46
0.990 (0.978–0.998)
LF/HF
9.83 ± 8.54
7.31 ± 7.96
0.919 (0.815–0.965)
Data represented as mean ± SD
HFnu (%)
18.29 ± 18.85
21.60 ± 18.22
0.971 (0.932–0.988)
ICC intraclass correlation coefficient, 95 % CI 95 % confidence interval
SD1 (ms)
3.87 ± 1.98
4.32 ± 1.38
0.948 (0.254–0.983)
SD2 (ms)
23.98 ± 14.80
24.32 ± 15.90
0.999 (0.992–1.000)
under estimate the LF parameter. As for the LF variable, the agreement between methods for the TP variable can be considered as acceptable (Table 2; Fig. 2). The LoAs for HF, LF/HF and HFnu were very large in both absolute and relative terms (Table 2; Fig. 2). All the power-spectrum related HRV parameters presented good ICCs (Table 3). However, in relation to the LoAs discussed above, we cannot conclude that the comparisons made for HF, LF/HF and HFnu parameters are favorable. SD1 presented wide LoAs for absolute and relative values (-0.02 to 0.70 ms, -0.50 to 17.54 %, respectively). In the case of SD2, the LoAs for absolute and relative values were tighter (-0.26 to 0.30 ms, -1.08 to 1.24 %, respectively). SD1 refers to the short-term variability [30] and is associated with the RMSSD and HF parameters, while SD2 is a measure of slow HRV [30]. In a study performed at rest [19], higher LoAs for SD1 than for SD2 were also found, in both the supine and standing positions. SD2 show an excellent low 95 % CI of ICC (0.992), although for SD1 was too low (0.254). These data are consistent with those observed in the concordance analysis. The results of the present study shows that the measurement error of the GOW system compared to the gold standard are broader in the case of the HRV parameters related to short time change (e.g., RMSSD, HF, and SD1). Therefore, rapid oscillations of HR showed worst LoAs whereas low oscillations seem to be more stable in the GOW system (Table 2). There are some important technical differences between both methods that can account for these differences. The sample rate of the GOW system is 250 Hz while ECG works at 1,000 Hz. Although we have down sampled the ECG data prior to RR detection, the low sample rate of the GOW system can be an important source of error measurement. Other difference between both methods is the R wave detection RR algorithm. Weippert et al. [21] have previously reported that it can be an important source of discrepancies, especially for the short-term change related HRV parameters. Moreover, the GOW system and the ECG had different electrodes
123
distribution. The GOW shirt has the textrodes disposed in both sides of the chest. None of ECG derivations correspond to this distribution. Therefore, the heart electrical activity was not recorded from the same location and this should add to the GOW error of measurement. The GOW system was designed to control training sessions of healthy subjects and to monitoring rehabilitation programs of cardiac patients. HRV analysis is particularly important in the case of cardiac patients since these data provides insight of the heart response to exercise and to the rehabilitation process [36, 37]. The results of the present study show that the measurement error of the GOW system is too large in order to accept the HRV data during exercise and therefore, do not meet the requirements as clinical tool. However, the GOW system does meet the requirements to controlling HR during exercise inasmuch as the agreement for the RR data and MeanRR variable is acceptable. In conclusion, the GOW system is a valid tool for controlling HR while performing physical activity, although its use as a clinical tool for HRV cannot be supported. Acknowledgments The authors thank Alicia Ricart for revising the level of English of this manuscript. Conflict of interest Javier Guillen was working in Weartech sl. at the time the research was conducted.
References 1. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology (1996) Heart rate variability: standards of measurement, physiological interpretation, and clinical use. Eur Heart J 17(3):354–381 2. Buchheit M, Gindre C (2006) Cardiac parasympathetic regulation: respective associations with cardiorespiratory fitness and training load. Am J Physiol Heart Circ Physiol 291(1):H451– H458. doi:10.1152/ajpheart.00008.2006 3. Zhang J, Tu W, Dai J, Lv Q, Yang X (2011) Dysfunction of preand post-operative cardiac autonomic nervous system in elderly
Australas Phys Eng Sci Med
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
patients with diabetes mellitus. Arch Gerontol Geriatr 53(3): 334–337 Carney RM, Blumenthal JA, Stein PK, Watkins L, Catellier D, Berkman LF, Czajkowski SM, O’Connor C, Stone PH, Freedland KE (2001) Depression, heart rate variability, and acute myocardial infarction. Circulation 104(17):2024–2028. doi:10.1161/ hc4201.097834 Jindal RD, Vasko RC, Jennings JR, Fasiczka AL, Thase ME, Reynolds CF (2008) Heart rate variability in depressed elderly. Am J Geriatr Psychiatry 16(11):861–866. doi:10.1097/JGP. 0b013e318180053d Karp E, Shiyovich A, Zahger D, Gilutz H, Grosbard A, Katz A (2009) Ultra-short-term heart rate variability for early risk stratification following acute ST-elevation myocardial infarction. Cardiology 114(4):275–283 Goldstein B, Toweill D, Lai S, Sonnenthal K, Kimberly B (1998) Uncoupling of the autonomic and cardiovascular systems in acute brain injury. Am J Physiol Regul Integr Comp Physiol 275(4):R1287–R1292 Biswas AK, Scott WA, Sommerauer JF, Luckett PM (2000) Heart rate variability after acute traumatic brain injury in children. Crit Care Med 28(12):3907–3912 Batchinsky AI, Cancio LC, Salinas J, Kuusela T, Cooke WH, Wang JJ, Boehme M, Convertino VA, Holcomb JB (2007) Prehospital loss of R-to-R interval complexity is associated with mortality in trauma patients. J Trauma 63(3):512–518. doi:10. 1097/TA.0b013e318142d2f0 Al Haddad H, Laursen PB, Chollet D, Ahmaidi S, Buchheit M (2011) Reliability of resting and postexercise heart rate measures. Int J Sports Med 32(08):598–605. doi:10.1055/s-00311275356 Arduini A, Gomez-Cabrera M-C, Romagnoli M (2011) Reliability of different models to assess heart rate recovery after submaximal bicycle exercise. J Sci Med Sport 14(4):352–357. doi:10.1016/j.jsams.2011.02.012 Kiviniemi A, Hautala A, Kinnunen H, Tulppo M (2007) Endurance training guided individually by daily heart rate variability measurements. Eur J Appl Physiol 101(6):743–751. doi:10.1007/ s00421-007-0552-2 Kiviniemi AM, Hautala AJ, Kinnunen H, Nissila¨ J, Virtanen P, Karjalainen J, Tulppo MP (2010) Daily exercise prescription on the basis of HR variability among men and women. Med Sci Sports Exerc 42(7):1355–1363. doi:10.1249/MSS.0b013e3181 cd5f39 Cottin F, Lepreˆtre PM, Lopes P, Papelier Y, Me´digue C, Billat V (2006) Assessment of ventilatory thresholds from heart rate variability in well-trained subjects during cycling. Int J Sports Med 27(12):959–967. doi:10.1055/s-2006-923849 Cottin F, Me´digue C, Lopes P, Lepreˆtre PM, Heubert R, Billat V (2007) Ventilatory thresholds assessment from heart rate variability during an incremental exhaustive running test. Int J Sports Med 28:287–294. doi:10.1055/s-2006-924355 Blain G, Meste O, Bouchard T, Bermon S (2005) Assessment of ventilatory thresholds during graded and maximal exercise test using time varying analysis of respiratory sinus arrhythmia. Br J Sport Med 39(7):448–452. doi:10.1136/bjsm.2004. 014134 Gamelin FX, Berthoin S, Bosquet L (2006) Validity of the Polar S810 heart rate monitor to measure R–R intervals at rest. Med Sci Sports Exerc 38(5):887–893. doi:10.1249/01.mss.0000218135. 79476.9c Gamelin FX, Baquet G, Berthoin S, Bosquet L (2008) Validity of the Polar S810 to measure R–R intervals in children. Int J Sports Med 29(2):134–138. doi:10.1055/s-2007-964995
19. Nunan D, Jakovljevic D, Donovan G, Hodges L, Sandercock G, Brodie D (2008) Levels of agreement for RR intervals and shortterm heart rate variability obtained from the Polar S810 and an alternative system. Eur J Appl Physiol 103(5):529–537. doi:10. 1007/s00421-008-0742-6 20. Walle´n M, Hasson D, Theorell T, Canlon B, Osika W (2012) Possibilities and limitations of the Polar RS800 in measuring heart rate variability at rest. Eur J Appl Physiol 112(3): 1153–1165. doi:10.1007/s00421-011-2079-9 21. Weippert M, Kumar M, Kreuzfeld S, Arndt D, Rieger A, Stoll R (2010) Comparison of three mobile devices for measuring R–R intervals and heart rate variability: Polar S810i, Suunto t6 and an ambulatory ECG system. Eur J Appl Physiol 109(4):779–786. doi:10.1007/s00421-010-1415-9 22. Kingsley M, Lewis MJ, Marson RE (2005) Comparison of Polar 810s and an ambulatory ECG system for RR interval measurement during progressive exercise. Int J Sports Med 26:39–44. doi:10.1055/s-2004-817878 23. Lymberis A, Paradiso R (2008) Smart fabrics and interactive textile enabling wearable personal applications: R&D state of the art and future challenges. Conf Proc IEEE Eng Med Biol Soc 2008:5270–5273 24. Curone D, Secco EL, Tognetti A, Loriga G, Dudnik G, Risatti M, Whyte R, Bonfiglio A, Magenes G (2010) Smart garments for emergency operators: the proeTEX project. IEEE Trans Inf Technol Biomed 14(3):694–701. doi:10.1109/titb.2010.2045003 25. Villalba E, Arredondo MT, Moreno A, Salvi D, Guillen S (2006) User interaction design and development of a heart failure management system based on wearable and information technologies. Conf Proc IEEE Eng Med Biol Soc 2006:400–403 26. Grant CC, Viljoen M, van Rensburg DCJ, Wood PS (2012) Heart rate variability assessment of the effect of physical training on autonomic cardiac control. Ann Noninvasive Electrocardiol 17(3):219–229. doi:10.1111/j.1542-474X.2012.00511.x 27. Niskanen J-P, Tarvainen MP, Ranta-aho PO, Karjalainen PA (2004) Software for advanced HRV analysis. Comput Methods Program Biomed 76(1):73–81. doi:10.1016/j.cmpb.2004.03.004 28. Tarvainen MP, Ranta-aho PO, Karjalainen PA (2002) An advanced detrending method with application to HRV analysis. IEEE Trans Biomed Eng 49(2):172–175. doi:10.1109/10.979357 29. Brennan M, Palaniswami M, Kamen P (2001) Do existing measures of Poincare plot geometry reflect nonlinear features of heart rate variability? IEEE Trans Biomed Eng 48(11):1342–1347. doi:10.1109/10.959330 30. Tulppo MP, Makikallio TH, Takala TE, Seppanen T, Huikuri HV (1996) Quantitative beat-to-beat analysis of heart rate dynamics during exercise. Am J Physiol Heart C 271(1):H244–H252 31. Carrasco M, Gaita´n R, Gonza´lez O, Ya´nez S (2001) Correlation among Poincare´ plot indexes and time and frequency domain measures of heart rate variability. J Med Eng Technol 25(6):240–248. doi:10.1080/03091900110086651 32. Bland JM, Altman DG (2007) Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 17(4):571–582 33. Pinna GD, Maestri R, Torunski A, Danilowicz-Szymanowicz L, Szwoch M, La Rovere MT, Raczak G (2007) Heart rate variability measures: a fresh look at reliability. Clin Sci 113(3): 131–140. doi:10.1042/cs20070055 34. Nunan D, Donovan G, Jakovljevic DG, Hodges LD, Sandercock GRH, Brodie DA (2009) Validity and reliability of short-term heart-rate variability from the Polar S810. Med Sci Sports Exerc 41(1):243–250 35. Bland JM, Altman D (1986) Statistical Methods for assessing agreement between two methods of clinical measurement. Lancet 327(8476):307–310
123
Australas Phys Eng Sci Med 36. Routledge FS, Campbell TS, McFetridge-Durdle JA, Bacon SL (2010) Improvements in heart rate variability with exercise therapy. Can J Cardiol 26(6):303–312
123
37. Rosenwinkel ET, Bloomfield DM, Arwady MA, Goldsmith RL (2001) Exercise and autonomic function in health and cardiovascular disease. Cardiol Clin 19(3):369–387
