Simple tai chi exercise for improving elderly postural ... - Springer Link

Artif Life Robotics (2015) 20:42–48 DOI 10.1007/s10015-014-0193-6

ORIGINAL ARTICLE

Simple tai chi exercise for improving elderly postural stability via complexity index analysis Cheng‑Wei Huang · Wei‑Hsin Chen · Heng‑Hui Chu · Bernard C. Jiang · Maysam Abbod · Jiann‑Shing Shieh

Received: 22 April 2014 / Accepted: 28 November 2014 / Published online: 7 January 2015 © ISAROB 2014

Abstract The main purpose of this study is to investigate twice a day simple 9-step tai chi effects of the center of pressure (COP) and physiological signals of elderly people. Data are collected from the COP signals, electromyography (EMG), and pulse oximetry for 1 min for the period of 12 weeks. The COP signals are analyzed using multivariate empirical mode decomposition and multivariate multiscale entropy to work out and compare the complexity index (CI). Subjects in this experiment are over 65 years old who are divided into 11 men and 7 women; the average age is 74 ± 8.18 years. In conclusion, it is found that tai chi

exercise can improve human body balance by just walking some simple steps in our experiment. However, we cannot find any effect or improvement in the pulse oximetry and EMG signals analysis. Keywords Center of pressure (COP) · Electromyography (EMG) · Multivariate empirical mode decomposition (MEMD) · Multivariate multiscale entropy (MMSE) · Complexity index (CI) · Six-minute walk test · Time-upand-go · Tai chi

1 Introduction This work was presented in part at the 19th International Symposium on Artificial Life and Robotics, Beppu, Oita, January 22–24, 2014. C.‑W. Huang · H.‑H. Chu · J.‑S. Shieh (*) Department of Mechanical Engineering, Yuan Ze University, Chung‑Li, Taiwan e-mail: [email protected] W.‑H. Chen Department of Industrial Engineering and Management, Yuan Ze University, Chung‑Li, Taiwan B. C. Jiang Department of Industrial Management, National Taiwan University of Science and Technology, Taipei, Taiwan M. Abbod School of Engineering and Design, Brunel University, London UB8 3PH, UK J.‑S. Shieh Innovation Center for Big Data and Digital Convergence, Yuan Ze University, Chung‑Li, Taiwan J.‑S. Shieh Center for Dynamical Biomarkers and Translational Medicine, National Central University, Chung‑Li, Taiwan

13

Falls are caused by many factors, including external factors such as environment and multiple drugs. The inherent risk factors such as age, a history of fall or recurring nearly fall without fall events, diseases, cognitive impairment, depression, sensory dysfunction, lower extremity muscle strength, balance and gait problems, the use of exercise walking aids, and daily living dysfunction. There are many studies concerned with the prevention of falling which are based on the multifactorial assessment tool to assess falling [1]. Elderly falling prevention intervention strategies includes fall risk assessment, multifactorial intervention, sport training, adjustment of medication, improved environment and vision, supplemental vitamin D and calcium, pacemaker and education. Balance exercise such as tai chi not can only reduce falling occurrences but also can get other physical and mental health benefits at the same time. Reducing elderly mental disorder treatment medication can significantly reduce falls, but cannot easily change the drug effects on the elderly. Environmental improvement intervention is suitable for the elderly who has fallen in the past. Cataract surgery can reduce most of the elderly’s bad fall caused by

Artif Life Robotics (2015) 20:42–48

vision problems. However, exercising tai chi can increase leg muscle strength and effectively improves lower extremity muscle strength. Tai chi exercise can also increase balance and reduce the risk of falling for the elderly [2]. Tai chi is one of Chinese traditional martial arts and a treasure of Chinese culture, it emphasizes on physical, mental and spiritual cultivation and practice. Tai chi is categorized as a recreational sport of low-middle intensity and aerobic, rhythmic, uncompetitive and whole body performing mode is suitable for various ages and can be practiced anytime, and anywhere. Center of pressure is an indirect measurement to measure Medial–Lateral (M–L) and Anterior–Posterior (A–P) [3] plane rocking postures. An effective nonlinear analysis method can effectively describe the posture stability and characteristics. The center of pressure (COP) signals are collected for body rocking posture using a force platform, then the displacement of center of pressure displacement is calculated for the Medial–Lateral and Anterior–Posterior directions [4]. Furthermore, the M–L and A–P displacements are analyzed for individual analysis. Therefore, the main purpose of this study is to investigate twice a day simple 9-step tai chi exercise effects on improving the center of pressure (COP) and physiological signals.

43

to define. The upper envelope is the cubic spline to link with the local maximum value of the continuous curve, and the lower envelope is the cubic spline to connect to the local minimum value of the continuous curve. Average upper and lower envelope is seen as a signal of a trend. The average envelope can be regarded as possible intrinsic mode functions. The possible intrinsic mode functions separation from the signal process is called sifting process. The intrinsic mode functions obtained from the sifting process must satisfy the following two conditions, otherwise it must run the sifting process again [6]. (1) In the whole time series, the number of all local extrema and zero-crossing difference cannot be more than one. (2) In any one time point, the average envelope must tend to zero. If these two conditions are satisfied, the separated signals called intrinsic mode functions (IMF) will be recorded as c1. Original signal and c1 subtraction can get residual signal. The residual signal is used as input to the decomposition of next intrinsic mode functions. Repeating this process can gradually be decomposed for different intrinsic mode functions until the residual signal is a monotonic function. Original signal after the n times decomposition can get different n groups’ IMFs signals. The relationship between decomposed by the empirical mode decomposition method signals and the original signal can be expressed by Eq. (1)

2 Method

X(t) = There are a few analysis algorithms, such as multivariate empirical mode decomposition (MEMD) and multivariate multiscale entropy (MMSE) that are used to analyze the center of pressure (COP), electromyography (EMG) data and other vital signs. 2.1 Analysis algorithms In this study, MEMD and MMSE are used to analyze the COP, EMG and other data. 2.1.1 Multivariate empirical mode decomposition The MEMD algorithm basically comes from empirical mode decomposition (EMD), which was proposed by Huang et al. [5], and has been widely used in nonlinear and non-stationary data analysis based on the inherent characteristics of the time series. EMD has groups of intrinsic mode functions (IMF) corresponding to the modal function within the system in different mechanism of reaction. The first step to calculate the EMD is to find the upper and lower envelope, which is based on the signal in the local maximum and local minimum value

n

i=1

ci + rn

(1)

where X(t) is the original signal, ci represents the number i IMF, rn is number n times residual signal. To diversely extend the application of EMD, MEMD method is being developed to decompose diversely nonlinear and non-stationary signals [7]. MEMD method not only overcomes the limit of a single EMD input, but also addresses the computing burden in calculating the noise signal residues in different channels after adding white noise. In addition, it is similar to EMD as a dyadic filter is used for multivariable input of each channel. Also it has the advantage of correction to align the corresponding IMFs from different channels to the same frequency range [8]. In the MEMD method, the average value of m(t) is calculated by multivariate-enveloped K direction vectors, as illustrated in Eq. (2)

m(t) = K −1

K

k=1

eθk (t)

(2)

where the {eθk (t)}K k=1 is a multivariable envelope in the vectors, along the K direction that predict multichannel input s(t) and multivariate IMF calculations via the s(t) − m(t) and stopping criteria. This process is repeated until all predicted standard signal is satisfied to the standard EMD stops.

13

44


Recently a study [6] has reported that the human body center of pressure signal is lower than 2 Hz. Accordingly, taking the Fourier transform for each IMF can indicate the frequency range of the original signal, therefore, a 2-Hz reference was chosen for analyzing the IMFs. 2.1.2 Multivariate multiscale entropy Entropy-based algorithms for measuring the complexity of physiologic time series have been widely used. These have proved to be useful in discriminating between healthy and disease patients [9]. Normally, healthy systems generate much more complex outputs than diseased ones. Traditional algorithms are single scale based and fail to account for the multiple time scales inherent in physiologic systems. Similar to EMD and MEMD, multivariate multiscale entropy (MMSE) is the evolution of multiscale entropy (MSE). MSE is developed by Costa et al. [10] and used to evaluate the complexity of signals over different time scales. Based on the approximate entropy, this method uses sample entropy to quantify the regularity of the finite length time series, which has been applied effectively in analysis of physiology, biology, and geosciences data [11–13]. Given a one-dimensional discrete time series {Xi, …, n X }, MSE first constructs multiple coarse-grained time series using the scale factor τ. Each element of the time series, {y(1)}, is according to Eq. (3) (τ )

yj

jτ =1 τ xi i=(j−1)τ +1

1 ≤ j ≤ N/τ .

(3)

For scale one, the time series{y(1)}is simply the original time series. The length of each coarse-grained time series is equal to the length of the original time series divided by the scale factor τ. The sample entropy is calculated for each coarse-grained time series plotted as a function of the scale factor τ [10]. Sample entropy reflects the conditional probability that two sequences of m consecutive data points which are similar to each other remain similar when one more consecutive point is included and being “similar” means that the value of a specific measure of distance is less than r [8]. The complexity degree of different combinations in each direction is measured in terms of the complexity index (CI), which is defined as the area under the MSE curve over all scales. MMSE calculates the relative complexity of the multichannel signals through the plot of the multivariate sample entropy. This makes it possible to assess structural complexity of multivariate physical or physiological systems, together with more degrees of freedom and enhanced rigor in the analysis. In the MMSE, if the multivariable sample entropy value is higher than most of the other scale,

13

multivariable time series is considered to be more complex than the others. This is the same with the original MSE [2]. 2.2 Experiment The purpose of this experiment is to improve elderly balance by performing nine-step tai chi exercise, and test whether it can improve the heart and lung function. This work is done in corporation with Taiwan Taoyuan Senior Citizens’ Home. Subjects in this experiment are over 65 years old who are 11 men and 7 women, and average age is 74 ± 8.18 years. The physical condition of these subjects is able to walk by themselves, do not need to be supported, and can stand on their own. 2.2.1 Measuring instruments To assess the improvement of body balance, the force platform measurement instrument developed in the previous research is being used. The system is named as Center of Pressure and Complexity Monitor System (CPCMS) (Fig. 1). The pressure measurement platform is designed to receive raw data of the balance signal. When the subject stands on the balance measurement device, the pressure measurement platform can receive raw COP signal. Pulse oximeter-measured oxygen saturation is a noninvasive method to monitor the oxygen saturation. This device is small, light-weight, reusable and portable as shown in Fig. 2a. The sensor is shown in Fig. 2b. This product is developed by Tatung Company. The device is connected to a computer to record SPO2 and heart rate. In this study the NeXus-10 Bluetooth Biofeedback (Fig. 3) is used to collect EMG signal. Figure 4 shows the

Fig. 1 Center of pressure and complexity monitor system (CPCMS)


experiment setup which shows the subject standing on the CPCMS and wearing Tatung pulse oximeter and NeXus-10 Bluetooth Biofeedback to collect all the data. 2.2.2 Experimental procedure

45

2.3 Data analysis The analysis procedure consists two parts. During the experiment, COP, SPO2 and EMG signals are recorded. SPO2 signal provides blood oxygen percentage, heart rate, where the respiration rate and psychological stress index

The experiment period for the nine-step tai chi exercise experiment is 12 weeks; conducted in the Taiwan Taoyuan Senior Citizens’ Home, data are collected every 2 weeks. Subjects do the exercise in the morning and evening every day. Rehabilitation division staff assists with the experiments. The experiment flowchart is shown in Fig. 5. The subjects are asked to stand on the CPCMS and connect the SPO2 and EMG device at the same time to measure COP, SPO2 and EMG. Each measurement lasts for 60 s, then 60 s break, then the measurement is repeated. The procedure is repeated more than three times. After finishing the procedure above, the subjects are asked to take 60 s break, then do the time-up-and-go test one time. The distance of walking is 3 meters. Take another 60 s break for next test, 6-min walk test where the distance is measured. Since all the subjects are elderlies, there is a person who helps beside them in case of any danger.

Fig. 4 Experiment schematic diagram. a Pulse-oximeter sensor, b CPCMS and c EMG connect to NeXus-10

Research group Explain experiment processes and sign subjects consents form

Fig.  2 a The Tatung pulse oximeter and b Tatung pulse-oximeter sensor

Subjects

Sign the form

Measure the COP, SPO2, EMG, time-up-andgo, and 6-minute walk be fore doing tai chi

Measure COP, SPO2 and EMG together Do time-up-and-go and 6-minute walk

Repeat the measurement procedure every two weeks for six times

End of one measurement Check if finish the whole experiment

NO

Do tai chi exercise everyday

YES End of whole experiment Fig. 3 NeXus-10 Bluetooth Biofeedback

Fig. 5 Experiment flowchart

13

46

Artif Life Robotics (2015) 20:42–48 COP Signal

Table 2 CI of MSE in ML

EMG signal

MSE_ML Resampling

ML/AP MEMD MSE/MMSE

nth

CI

1 2 3 4 5

5.651 ± 2.22 5.458 ± 1.842 5.629 ± 2.572 6.288 ± 2.678 5.823 ± 2.171

1 vs 2 1 vs 3 1 vs 4 1 vs 5

0.704 0.971 0.265 0.691

6

6.687 ± 2.122

1 vs 6

0.079

p value

1 vs 6 represents a comparison of 1st with 6th CI Result and discussion

Fig. 6 COP and EMG analysis process flowchart

(PSI) are inferred. From the result, the second part involves analyzing the tai chi exercise affect on the subjects’ heart and lung function. The COP and EMG analysis process is shown in Fig. 6. Human body center of pressure can be used to assess personal balance mechanism. Since the human body center of pressure signal frequency is very low, the force platform sample rate is 500 Hz and so the signal is resampled at 50 Hz. After the calculation, M–L and A–P signals are acquired. MEMD is used to filter out noise and then use MMSE to calculate entropy value. Finally, the complexity index (CI) is calculated to assess the difference in human body balance ability due to the tai chi exercise.

3 Analysis result The COP displacements in two directions of A–P and the M–L are often used to characterize the COP stabilogram. Before performing the MSE and MMSE analysis, the IMF selection has to be made. MEMD is performed first, then the Fourier transform for each IMF to get the frequency range. The frequencies for each IMF are shown in Table 1.

Table 1 IMF frequencies CPCMS IMF 2

IMF 3

IMF 5

IMF 6

X

9.17 ± 1.71

4.22 ± 0.84

1.24 ± 0.26

0.73 ± 0.18

Y

9.53 ± 1.59

4.54 ± 1.04

1.22 ± 0.24

0.73 ± 0.15

13

IMF 5+6 are selected since these indicate less than 2 Hz signals. Tables 2 and 3 show the results of MSE of AP and ML. From the results, it is obvious that CI of the 6th measurement is higher than the 1st measurement in both AP and ML directions. MMSE is also used to combine both ML and AP directions into one result. Table 4 shows the CI value of the MMSE. From the result, it can be seen that CI of 6th is higher than 1st. From the time-up-and-go test, the subjects need to stand up from a chair and walk to a 3-mfar target then walk back to sit down. The shorter finish

Table 3 CI of MSE in AP MSE_AP nth

CI

1 2 3 4 5

7.386 ± 2.406 8.38 ± 2.822 7.492 ± 2.797 8.973 ± 3.679 8.331 ± 2.253

1 vs 2 1 vs 3 1 vs 4 1 vs 5

0.134 0.885 0.072 0.108

6

9.318 ± 2.397

1 vs 6

0.020

p value

1 vs 6 represents a comparison of 1st with 6th

Table 4 CI of MMSE MMSE nth

CI

1 2 3 4 5

13.095 ± 1.887 14.687 ± 1.639 14.042 ± 1.559 14.409 ± 1.93 13.858 ± 1.557

1 vs 2 1 vs 3 1 vs 4 1 vs 5

0.003 0.132 0.013 0.105

6

14.577 ± 1.433

1 vs 6

0.006

p value



47

Table 5 Time-up-and-go result

time means the speed and stability of walking increase. Table 5 indicates that the time to finish the procedure is shorter after 6th time measurement. The same trend can be seen for the time-up-and-go test, the longer distance is the better, as illustrated in Table 6. Tables 7 and 8 show the analysis results for pulse oximetry. The heart rate, blood oxygen concentration and respiration rate are extracted from the pulse oximetry. These results indicate no regular changes or improvements in these four parts. The p values for before and after doing the tai chi exercise are bigger than 0.05.

Time-up-and-go nth

Time (s)

1 2 3 4 5

16.195 ± 6.124 15.54 ± 4.42 14.78 ± 3.97 14.1 ± 3.27 14.45 ± 3.25

1 vs 2 1 vs 3 1 vs 4 1 vs 5

0.599 0.216 0.184 0.241

6

13.72 ± 3.14

1 vs 6

0.060

p value


4 Conclusion and discussion Table 6 6-minute walk test

The MSE result of both COP direction and the MMSE combination method shows that the tai chi exercise can improve human body balance even when just walking few simple steps. However, in this experiment no improvement in the pulse oximetry and EMG signals analysis is found. Furthermore, more rigorous study can be conducted to investigate the relation between complexity of the COP data, physical ability and tai chi. In the future, this experiment can be extended to elderlies with diseases, and then construct an evaluation index for home care for the healthy or diseased elderlies.

6-Min walk nth

Distance (m)

1 2 3 4 5

262.756 ± 61.603 272.806 ± 67.3 289.243 ± 70.513 297.52 ± 70.513 295.698 ± 75.078

1 vs 2 1 vs 3 1 vs 4 1 vs 5

0.234 0.011 0.000 0.005

6

305.052 ± 72.974

1 vs 6

0.000

p value


Table 7 Pulse oximetry analysis result for heart rate

HR 1

2

3

4

5

6

1 2 3 4 5 6 7 8

88.4 72.13 85.09 58.71 69.94 92.42 66.36 77.61

79.65 71.03 90.35 60.29 76.76 85.69 69.73 78.1

79.66 71.03 86.51 107.31 78.87 73.77 62.84 84.6

87.69 72.39 99.24 42 65.5 75.1 70.52 83.08

74.84 74.99 83.78 81.74 64.38 73.92 65.93 97.19

90.2 73.99 94 50.33 66.67 75.45 72.03 76.94

9 10 11 12 13 14 15 16 17 18 Mean

19.87 78.1 88.08 84.6 59.39 67.53 65.86 60.13 71.57 86.75 71.81

85.84 85.34 85.84 85.41 59.2 83.54 59.66 58.49 75.81 81.25 76.22

86.65 84.6 86.65 81.66 72.07 87.29 53.45 55.11 76.04 87.02 78.62

88.55 83.08 88.55 77.08 79.51 98.8 58.94 64.09 69.13 81.56 76.93

89 77.34 81.18 90.01 74.86 80.75 55.64 67.24 73.99 39.86 74.81

92.98 73.1 70.48 90.14 69.37 72.46 59.46 55.06 73.61 90.36 74.81

SD

16.89

10.68

12.85

14.21

13.24

12.78

p value

0.272

0.425

0.685

0.588

1.000

13

48


Table 8 Pulse oximetry analysis result for blood oxygen concentration

SPO2 1

2

3

4

5

6

1 2 3 4

96.92 94.34 96.04 90.68

95.41 64.43 94.14 99

95.41 64.45 96.39 90.94

96.73 97.04 95.13 97.56

99 95.85 94.55 99

97.37 96.56 92.45 99

5 6 7 8 9 10 11 12 13 14 15 16 17 18 Mean SD

96.24 97.69 99 91.43 99 99 98.34 99 98.07 98.93 98.96 98.32 95.94 94.36 96.79 2.61

96.18 97.15 99 99 99 99 99 91.13 97.16 95.87 93.05 95.08 92.68 94.43 94.48 7.92

95.25 95.13 93.89 99 99 99 99 91.45 94.15 93.86 93.91 98.28 92.74 97.74 93.87 7.8

81.99 95.36 93 96.78 95.42 96.78 95.42 94.24 94.65 96.79 96.56 95.65 97.43 97.07 95.2 3.51

90.94 96.21 98.44 98.24 94.9 98.06 83.55 94.73 95.53 93.64 95.65 97.3 98.25 98.64 95.69 3.73

94.7 99 99 98.24 96.37 98.97 83.55 93.76 96.59 92.61 96.3 96.57 99 96.84 95.94 3.75

p value

Acknowledgments This research is supported by Ministry of Science and Technology in Taiwan through Grant number of NSC1022221-E-155-028-MY3. This research is also supported by the Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan which is sponsored by Ministry of Science and Technology (NSC102-2911-I-008-001).

References 1. World Health Organization (2007) WHO global report on falls prevention in older age, World Health Organization 2. Ahmed MU, Mandic DP (2012) Multivariate multiscale entropy analysis. IEEE Signal Process Lett 19(2):91–94 3. Sezer O, Ferdjallah M (2005) Adaptive autoregressive model for the analysis of center of pressure in healthy subjects during quiet standing. In: IEEE 48th Midwest Symposium on Circuits and Systems, vol. 1. Covington, pp 495–498 4. Snoussi H, Amoud H, Doussot M, Hewson D, Duchene J (2006) Reconstructed phase spaces of intrinsic mode functions. In: Application to Postural Stability Analysis, Engineering in Medicine and Biology Society. EMBS ‘06. 28th Annual International Conference of the IEEE, New York, pp 4584–4589 5. Huang NE, Shen Z, Long SR, Wu MC, Shih HH, Zheng Q, Yen NC, Tung CC, Liu HH (1998) The empirical mode decomposition

13

0.232

0.355

0.533

0.611

0.506

and the Hilbert spectrum for nonlinear and nonstationary time series analysis. Proc R Soc Lond 454:903–995 6. Freitas SMSF, Wieczorek SA, Marchetti PH, Duarte M (2005) Age-related changes in human postural control of prolonged standing. Gait Posture 22:322–330 7. Rehman N, Mandic DP (2010) Multivariate empirical mode decomposition, in Proc. Roy Soc A 466:1291–1302 8. Rehman N, Mandic DP (2010) Filter bank property of multivariate empirical mode decomposition, IEEE Trans. Signal Proc 59:2421–2426 9. Costa M, Goldberger AL, Peng CK (2002) Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett 89:068102 10. Costa M, Goldberger AL, Peng CK (2005) Multiscale entropy analysis of biological signals. Phys Rev E 71:021906-1–02190618 11. Thuraisingham RA, Gottwald GA (2006) On multiscale entropy analysis for physiological data. Phys A 366:323–332 12. Liu Q, Wei Q, Fan SZ, Lu CW, Lin TY, Abbod MF, Shieh JS (2012) Adaptive computation of multiscale entropy and its application in EEG signals for monitoring depth of anesthesia during surgery. Entropy 14(6):978–992. doi:10.3390/e14060978 13. Costa M, Healey JA (2003) Multiscale entropy analysis of complex heart rate dynamics: discrimination of age and heart failure effects, computers in cardiology 30:705–708