Gait & Posture 40 (2014) 581–586
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Complexity of human postural control in subjects with unilateral peripheral vestibular hypofunction Jia-Rong Yeh a,b, Men-Tzung Lo a,b, Fu-Ling Chang c,d, Li-Chi Hsu e,f,* a
Research Center for Adaptive Data Analysis, National Central University, Taoyuan, Taiwan Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taoyuan, Taiwan c Taiwan Textile Research Institute, New Taipei City, Taiwan d National Yunlin University of Science and Technology, Graduate School of Design, Yunlin, Taiwan e Department of Neurology, Taipei Veterans General Hospital, Taipei, Taiwan f National Yang-Ming University School of Medicine, Taipei, Taiwan b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 20 January 2014 Received in revised form 16 April 2014 Accepted 30 June 2014
Complexity is a new measure for identifying the adaptability of a complex system to meet possible challenges. For a center of pressure (COP) time series, the complexity measure represents the stability of postural control. In this study, multiscale entropy (MSE) was used to evaluate the complexity of COP time series in six test conditions of sensory organization test (SOT). Complexity index (CI) is defined as the summation of entropies with coarse-graining scales 1–20 by MSE. A total of 51 subjects belonging to 3 groups – healthy-young, healthy-elderly and dizzy – were recruited in this study. The COP signals in both anteroposterior (AP) and mediolateral (ML) directions were analyzed respectively. According to our results, the CI of AP-direction COP time series is significantly correlated to the equilibrium score, which represents the stability of postural control in SOT. The AP-direction sway is significant larger than the ML-direction sway, particularly in the test conditions with sway-surface. In additions, the CI of APdirection COP for the healthy-elderly and dizzy groups are significantly lower than those for the healthy young group in the test conditions 1–4. The CI of ML-direction COP for the healthy-elderly group is significantly lower than those for the healthy-young and dizzy groups under test conditions 3 and 6. These results show that the complexity loss is a common status of AP-direction COP time series for both healthy-elderly and dizzy groups, and the complexity of ML-direction COP time series for subjects with unilateral vestibular dysfunction is higher than that for the healthy-elderly group specifically under test conditions 3 and 6. ß 2014 Elsevier B.V. All rights reserved.
Keywords: Complexity Center of pressure Multiscale entropy Sensory organization test Unilateral vestibular dysfunction
1. Introduction It is well-known that the integration of sensory systems, especially the visual, somatosensory and vestibular systems, are in close relation to the patients’ postural control [1]. In addition to direct observation of postural stability, posturography is a wellknown method to study balance disorders. The sensory organization test (SOT) paradigm of the posturographic test is a common protocol used in clinical practice. The principle of SOT evaluation is to selectively disrupt the support surface and/or the visual input to
* Corresponding author at: Department of Neurology, Taipei Veterans General Hospital, No. 201, Sec. 2, Shih-Pai Rd., Taipei 11217, Taiwan. Tel.: +886 2 28757578. E-mail addresses: [email protected], [email protected] (L.-C. Hsu). http://dx.doi.org/10.1016/j.gaitpost.2014.06.016 0966-6362/ß 2014 Elsevier B.V. All rights reserved.
measure the subject’s ability to use the remaining sensory inputs for postural control [2]. As the degree of task difficulty increased from open-eyes to closed-eyes and support surface from rigidsurface to sway-referencing surface, the instability of posture increased as well [3]. When analyzed properly, the SOT can determine which sensory systems (visual, somatosensory or vestibular) the patient relies upon most to maintain stability. In clinical practice, a formal SOT report includes the composite equilibrium scores in six test conditions and four ratios of sensory analysis. The four ratios are considered to reflect the functional conditions of somatosensory (SOM), visual (VIS), vestibular (VEST) and visual preference (PREF). In addition to the classic SOT report, the trajectory of the center of pressure (COP) time series can be measured in both anteroposterior (AP) and mediolateral (ML) directions respectively for all trials of SOT. In this study, COP time series in both AP and ML directions in six test conditions were
582
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
analyzed for characterizing the dynamical features of postural sway with disrupted and/or impaired sensory inputs in balance disorder patients. Aging and sensory disorders are important factors that affect the ability of postural control in human. The decrease of complexity of postural control system was proposed as a key signature of aging and sensory disorders [4–6]. In previous studies of complexity analysis, various algorithms have been proposed to estimate information entropy, i.e., complexity [7–10]. Multiscale entropy (MSE) analysis is based on the definition of the sample entropy (SampEn) proposed by Richman and Moorman [8]. Costa et al. defined a new measurement of ‘complexity index (CI)’ as the summation of the values of SampEn on multiple time scales [9]. The CI of both ML- and AP-direction COP time series represents the dynamical features of responses of postural control. So far, the CI acts as a good measurement of complexity in many applications [11–14]. In this study, the CI derived by the MSE method was used to evaluate the complexity features of COP time series in different test conditions of SOT for a total of 51 subjects with differences in age and functional conditions of vestibular systems. According to our results, the CI of COP time series is significantly correlated with the stability of postural control. A significant correlation between CI and equilibrium score (EQS) is identified by Pearson’s correlation coefficient with a value of 0.70. Moreover, the complexity of AP sway is negatively correlated to age in test conditions 1–4. The difference between the complexity of APdirection COP time series for the healthy elderly (HE) and dizzy (DZ) group is not significant. On the other hand, complexity of MLdirection COP time series represents as a critical assessment for identifying the dynamic property of postural control for the DZ group in test conditions 3 and 6, which shows significantly differences in comparisons with that of the HE group. 2. Methodology 2.1. Subjects The study was conducted at the Taipei Veterans’ General Hospital, which is a tertiary referral hospital in northern Taipei, Taiwan. The 51 subjects were divided into three groups: a healthy young (HY) group (19 males, 4 females, mean age: 36.2 5.7, range: 24–48), a healthy elderly (HE) group (3 males, 6 females, mean age: 68.6 5.2, range: 62–75) and a dizzy/imbalance (DZ) group (10 males, 9 females, mean age: 66.7 13.4, range: 40–85). Healthy subjects (HY and HE) had no past history of peripheral or central vestibular disorders, other medical or neurological diseases known to cause dizziness or gait disturbance and normal vestibular function as assessed by classical clinical otoneurologic tests including videooculographic and caloric tests. DZ patients all complained of chronic (>1 month) dizziness and unsteadiness. In addition, they all had unilateral peripheral vestibular hypofunction as documented by the caloric test (unilateral canal paresis 325%). Otherwise, none of the patients had any central lesions that could be interpreted as the origin of their dizziness or disequilibrium after appropriate clinical and/or neuroradiological examinations. The study hospital’s Institutional Review Board approved this trial and informed consent was obtained from each participant. 2.2. Sensory organization test (SOT) Computerized dynamic posturography is a functional test of the visual, vestibular and proprioceptive systems’ contribution to the maintenance of upright posture. The testing provides patients with a combination of present, absent, or distorted visual or proprioceptive cues based on a movable platform support and visual
surrounding. In the present study, the SMART Balance Master (NeuroCom, Clakamas, OR, USA) platform posturography was used for evaluation. The SOT protocol objectively identifies abnormalities in the patient’s use of the three sensory systems that contribute to postural control: somatosensory, visual and vestibular. Six test conditions for the SOT were designed: (1) eyes open with fixed surrounding and support; (2) eyes closed with fixed support; (3) sway-referenced surrounding with fixed support; (4) eyes open with fixed surrounding and sway-referenced support; (5) eyes closed with sway-referenced support; (6) sway-referenced surroundings and support. Equilibrium score (EQS) were measured as a nondimensional percentage comparing individual’s peak amplitude of AP sway with limit of stability (LOS). The sensory analysis aims to interpret the functional compensations in sensory systems using ratios of EQS in different test conditions. Each subject undertook 3 successful trials to keep standing without falling under 6 test conditions. If a subject fell in test, he was asked to take a break and one more trial. The COP time series was derived from the raw recordings of the force platform at a sampling rate of 100 Hz. 2.3. Multiscale entropy (MSE) The MSE method aims to measure the degree of complexity of a time series on multiple time scales. CI is defined as the summation of entropies for coarse-grained time series with scales of 1 n. The entropy of a coarse-grained time series is evaluated by previously suggested method so-called sample entropy (SampEn) [8]. The basic idea of the definition of SampEn is to provide a measure of ‘predictability’ in a time series by testing if there are any repeated patterns of various lengths. To calculate the predictability in a time series with N data points, a sequence of m successive points is selected as a vector from all possible m-point vectors, i.e., a total of N m + 1. Then, all m-point vectors that are similar to the selected one are counted, excluding the selected on itself. The number of similar vectors is denoted as nr;m where the index i is denotes the i vector number and r is the similarity tolerance. A difference between two points smaller than similarity tolerance is denoted as two points are similar to each other. All points in a vector are similar to the corresponding points in another vector, then two vectors are similar. The ratio nr;m =ðNm þ 1Þ represents the i probability of finding an m-point vector similar to the selected one. The average of all probabilities, which is computed for every possible m-point vector in the time series, represents the probability that two randomly selected vectors are similar to each other. The successive point of an m-point vector from the time series added to an m-point vector to result a (m + 1)-point vector. Considering two vectors that are similar to each other, the two resulting (m + 1)-point vectors may or may not be similar to each other. As if the resulting (m + 1)-point vectors that are similar to each other, we then say the (m + 1)th point is predictable when the two m-point vector are similar to each other. Therefore, the predictability, denoted as Cr,m, can be defined as: C
r;m
¼
PNm
r;mþ1 i¼1 ni P Nmþ1 r;m 1 ni i¼1 ðNmþ1Þ 1 ðNmÞ
(1)
Theoretically, predictability represents the regularity and entropy represents the irregularity of a time series. Therefore, sample entropy (SampEn) can be defined as the natural logarithm of the inverse of Cr,m. This definition of SampEn was extended by Costa et al. [9] to define the MSE as a result of successive coarsegraining of the time series. The coarse-graining was done by averaging the data points in given non-overlapping windows, t is
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
the number of data length of window, called the coarse-graining scale. The entropies versus their coarse-graining scales can be shown as a curve in a MSE plot. The under-curve area represents the summations of the entropies on multiple coarse-graining scales as a new assessment denoted as complexity index (CI). 2.4. Statistical analysis The population distributions of the EQS and CI were more Poisson-like than Gaussian-like. Therefore, the nonparametric Kolmogorov–Smirnov test (K–S test) was used to verify the difference between the two populations. The level of significance for the K–S test was set at P < 0.05. 3. Results 3.1. Results of sensory organization test A classical routine SOT report contains EQS in six test conditions and four ratios of sensory analysis. Table 1 shows the results of SOT report. The age of the HY groups is significantly lower than the other two groups. The difference between the HE and DZ groups is insignificant. As shown in Table 1, the EQS of test 1 in HE group was significantly lower than those for the other two groups which mean poorer balance stability during quiet stance in this group. Moreover, the EQS of tests 4 and 5 in HY group were significantly higher than those of the other two groups. These imply that the balance stability of the HY group is better than the other groups. Based on subjects’ ages and vestibular function status, aging might be a contributing factor of the difference between the HY and HE groups and the difference between the HE and DZ group might be attributed to the dysfunction of vestibular system. 3.2. Results of complexity analysis by multiscale entropy (MSE)
[(Fig._1)TD$IG]
The COP time series in both AP and ML directions were analyzed by MSE method respectively. The MSE plots for the three groups in six test conditions are shown in Figs. 1 and 2. Fig. 1 shows the MSE plots of ML-direction COP time series and Fig. 2 shows those for AP-direction COP time series. The CI was defined as the summation of entropies on the first 20 coarse-graining scales in this study. Therefore, the area under an MSE plot represents the complexity of a COP time series in a specific test condition. As shown in Figs. 1 and 2, the complexity of both ML- and AP-direction COP time series for the HY group is the highest one among three groups in all test conditions. In test condition 1, the subjects were instructed to stand on the fixed force plate with eyes opened. No sensory inputs of postural control are disrupted in test condition 1, so the complexity of COP time series is the highest. Whereas anyone of the sensory input is disrupted in the other test conditions, the complexity of COP time series is declined. On the other
583
Table 1 Statistical results of the original SOT reports comprising equilibrium scores (EQS) of six test conditions and four ratios of sensory analysis. The values were shown in mean standard deviation. Parameter
Healthy young
Dizzy
Healthy elderly
Subject number Age EQS of C1 EQS of C2 EQS of C3 EQS of C4 EQS of C5 EQS of C6 SOM VIS VEST PREF
23 36.2 5.7dd,ee 95.67 1.15ee 92.06 1.96 91.87 3.89e 87.28 4.58d,e 67.65 9.03d,ee 70.39 13.80d .962 .017 .912 .047d .707 .093dd,e 1.018 .086
19 68.6 5.2yy 95.44 1.35ee 92.18 3.01 89.33 4.83 82.07 6.35y 56.56 8.42y 55.07 15.23y .966 .026 .860 .060y .593 .089yy .972 .100
9 66.7 13.4yy 93.33 1.59yy,dd 90.54 2.16 87.38 5.90y 82.08 6.77y 56.25 14.32yy 62.96 9.74 .970 .017 .879 .064 .601 .145y 1.027 .048
Mark of ‘y’ means statistically significant difference in comparing with healthy young. Mark of ‘d’ means statistically significant difference in comparing with dizzy. Mark of ‘e’ means statistically significant difference in comparing with healthy elderly. Single character means with P-value < 0.05; twin characters means P-value < 0.01. hand, the complexities of AP-direction COP time series are significantly lower than those of ML-direction COP time series by comparing the results shown in Figs. 1 and 2. Fig. 3 depicts the results of CI for three groups in SOT. As shown in Fig. 3, the declines of complexity are strongly correlated to the difficulties of postural control in various test conditions. Sway-referenced surface causes the most significant complexity decline in comparing the CI values in test conditions of 1 and 4. The increasing instabilities of postural control caused by different sensory disruptions can be also double checked in sensory analysis. In addition, a significant correlation between EQS and the CI of AP-direction COP time series is verified by Pearson’s correlation coefficient with value of 0.6995. The correlation coefficient between EQS and CI of ML-direction COP time series is 0.6620. Therefore, CI might be considered a good assessment tool for the evaluation of balance stability. Furthermore, the complexity features of ML-direction COP time series are significantly different from those of AP-direction COP time series. The CI of APdirection COP time series in the HY group are significantly higher than those of the HE and DZ groups in test conditions 1–4. These results imply that aging is responsible for the reduction of complexity in AP-direction COP time series. The differences between the CI of AP-direction COP time series of HE and DZ groups in all of the six conditions of SOT are insignificant. In summary, the complexity decline of AP-direction time series is significant for elderly subjects regardless of the presence or absence of unilateral vestibular dysfunction.
Fig. 1. The MSE plots for the ML direction COP time series under six SOT test conditions.
[(Fig._2)TD$IG]
584
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
Fig. 2. The MSE plots for the AP direction COP time series under six SOT test conditions.
On the other hand, the CI values of ML-direction COP time series represent different dynamical features in comparison with those of AP-direction COP time series. The CI values of ML-direction COP time series for the HY group is the highest one and those for the HE group is the lowest one among the three groups. These results delineate that the complexity decline in the ML-direction COP time series for the DZ group is less than that for the HE group. Of note, there are significant differences between the CI values of ML-direction COP time series for the HE and DZ groups in test conditions 3 and 6. This finding implies that the CI of ML-direction COP time series in test conditions 3 and 6 might potentially reflect the specific features of posturographic evaluation for subjects with unilateral vestibular pathology.
4. Discussion
[(Fig._3)TD$IG]
In this study, the CI calculated by the MSE method was used to be a new measurement for assessing the postural control mechanism. As a general hypothesis, a system with high complexity reflects high capability to adapt to a constantly changing environment. Therefore, a higher CI value represents a better ability to maintain postural stability. Both EQS and CI are good tools for the assessment of balance ability of postural control. However,
the EQS is measured as the ratio between individual’s peak amplitude of AP sway and limit of stability, which represents a linear parameter directly associated with the magnitude of AP sway. Different from the linear parameter of EQS, the CI is a nonlinear parameter associated with the dynamical property of COP time series, which is independent from the magnitude of postural sway. A decrease of balance stability may cause an increase of postural sway magnitude and reduce the complexity of COP time series simultaneously. The changes of sway magnitude are essentially different from the changes of complexity of COP time series. Therefore, the performances of these two parameters should be different in posturography analysis and the analysis results should be compensative to each other. In this investigation, we have proved that the complexity of COP time series in specific test conditions of SOT actually reflects the differences in posturographic performance between the groups with and without unilateral vestibular dysfunction, which cannot be verified by the parameters based on linear assessment of EQS.
Fig. 3. The statistical results of the complexity index (CI) for three groups under six test conditions of SOT. The data is shown in mean values and standard errors.
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586 Table 2 Statistical results of complexity index (CI) values for AP- and ML-direction CoP time series. The values were shown in mean standard deviation. Healthy young For ML-direction of CoP time series CI in test 1 19.87 4.49e CI in test 2 20.46 4.55dd,ee CI in test 3 19.04 4.87e CI in test 4 18.87 3.62e CI in test 5 17.57 4.42d,e CI in test 6 17.83 4.44ee For AP-direction of CoP time series CI in test 1 18.07 4.05dd,ee CI in test 2 18.03 4.48d,ee CI in test 3 17.02 4.80d,ee CI in test 4 17.81 4.57dd,ee CI in test 5 13.97 5.49 CI in test 6 15.23 5.89e
Dizzy
Healthy elderly
19.18 4.48 18.14 5.68yy 18.74 4.86ee 17.98 4.54 16.14 4.58y 16.86 4.61e
17.67 3.59y 15.53 3.88yy 16.31 4.48y,dd 16.20 4.69y 15.30 3.76y 14.20 3.87yy,d
15.93 5.00yy 15.42 5.34y 13.87 5.25y 13.97 5.54yy 13.41 5.66 13.86 6.20
15.34 3.88yy 15.21 4.67yy 13.86 4.46yy 14.06 4.33yy 13.42 4.27 12.59 4.20y
Mark of ‘y’ means statistically significant difference in comparing with healthy young. Mark of ‘d’ means statistically significant difference in comparing with dizzy. Mark of ‘e’ means statistically significant difference in comparing with healthy elderly. Single character means with P-value < 0.05; twin characters means P-value < 0.01.
Moreover, most of the subjects with unilateral vestibular dysfunction are elderly in this study. Therefore, we also recruited healthy elderly subjects with age distribution similar to that of the subjects with unilateral vestibular dysfunction. So, the difference between the HE and DZ groups might represent the difference caused by the vestibular dysfunction. As shown in Table 1, only the EQS in test condition 1 revealed significant difference between HE and DZ groups. It is interesting that the EQS in test condition 1 for the HY group is insignificantly different from that for the DZ group. It implies that subjects with chromic unilateral peripheral vestibular dysfunction might have good compensation of postural controls in stress-free stance based on the results of EQS. The compensatory effect of the normal side of inner ear might keep the stability of postural control in the test condition of stress-free stance. On the other hand, the CI of AP-direction COP time series also does not differentiate those changes due to aging process from those caused by the unilateral peripheral vestibular dysfunction as shown in Table 2. The CI values of AP-direction COP time series show that aging causes the decline of complexity in AP-direction sway in test conditions 1–4. The complexity of AP-direction COP time series for the HY group is significantly higher than those for HE and DZ groups. On the contrary, the CI value of ML-direction COP time series presents different results in classifications among these three groups. In test conditions 3 and 6, the CI value of MLdirection COP time series for the DZ group is significantly higher than that for the HE groups. This result implies that subjects with unilateral vestibular dysfunction had better balance ability for maintaining ML-direction sway than healthy elderly subjects in the test conditions with sway-referenced visual surrounding. As the report in a previous study, the effects of visual stimulation on balance, such as test conditions 3 and 6, in patients with unilateral vestibulopathy reflect the type of sensory deficit and can be considered to be specific to such a deficit [14]. According to our observations reported in a previous study [15], sway referencing in the AP direction caused increased low-frequency fluctuations of AP-direction COP time series for both HE and DZ groups. However, changes of low-frequency fluctuations in ML-direction were insignificant for the DZ group. This implies that sway referencing in AP direction caused the increased low-frequency fluctuations in both directions of AP and ML for the HE group, but only in AP direction for the DZ group. The increased low-frequency fluctuation can cause the increased standard deviation of COP time series
585
and the decreased CI in a complexity analysis by MSE. Thus, CI in AP direction for the DZ group is similar to that for the HE group, and CI in ML direction for the DZ group is significantly higher than that for the HE group. In addition, visual input was disturbed by eye closing and both somatosensory and vestibular inputs were disturbed by sway supporting in AP direction for the test condition of SOT 5. There are no significant differences in sensory inputs among three groups for test condition of SOT 5. It resulted in similar AP-direction MSE plots and CI values for three groups in test condition of SOT 5. Moreover, why the elderly subjects performed even less than the DZ group? Decline in the integrity and functionality of sensory systems associated with advancing age [16,17] might play an important role because this may lead to postural control deficit in the elderly. Further studies are needed to test the role of conflict visual input in the maintenance of postural stability in the elderly. Nonetheless, our finding implies that the compensation of unilateral vestibular input drove a significant increase of ML-direction complexity as a specific pattern in posturography. In summary, the technique of complexity analysis was used to investigate the dynamic properties of COP time series in different test conditions of SOT. The nonlinear parameters derived by complexity analysis present different classification results in comparison with linear parameters. We suggest that the nonlinear complexity analysis is valuable for investigating the properties of a dynamic system with multiple control mechanisms. A welldesigned test protocol and an appropriate analysis algorithm will benefit to detail the functional conditions of a complicated system, such as the postural control system. Acknowledgements The authors wish to express their thanks for supports from the National Science Council (Taiwan) (Grant No 99-2627-B-008-003, and 97-2314-B-038-002-MY3), for joint funding from National Central University (Grant Nos CNJRF-99CGH-NCU-A3 and VGHUST100-G1-4-3), NSC support for the Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan (NSC 99-2911-I-008-100); and grants from the Taipei Veterans General Hospital (V99C1-194, V100C-205 and V101C-200). Conflict of interest statement. The authors disclose there are no potential conflicts of interest with other people or organizations in this study.
References [1] Woolacott MH, Shumway-Cook A. Attention and the control of posture and gait: a review of an emerging area of research. Gait Posture 2002;16:1–14. [2] Nashner LM. Practical management for the dizzy patient: computerized dynamic posturography. Philadelphia, PA: Lippincott Williams & Wilkins; 2008. [3] Negahban H, Salavati M, Mazaheri M, Sanjari AM, Hadian MR, Parnianpour M. Non-linear dynamical features of center of pressure extracted by recurrence quantification analysis in people with unilateral anterior cruciate ligament injury. Gait Posture 2010;31:450–5. [4] Goldberger AL, Amaral LA, Hausdorff JM, Ivanov P, Peng CK, Stanley HE. Fractal dynamics in physiology: alternations with disease and aging. Proc Natl Acad Sci USA 2002;99:2466–71. [5] Goldberger AL, Peng CK, Lipsitz LA. What is physiologic complexity and how does it change with aging and disease? Neurobiol Aging 2002;23:23–6. [6] Lipsitz LA. Dynamics of stability: the physiologic basis of functional health and frailty. J Gerontol A: Biol Sci Med Sci 2002;57:B115–25. [7] Pincus SM. Approximate entropy as a measure of system complexity. Proc Natl Acad Sci USA 1991;88:2297–301. [8] Richman JS, Moorman JR. Physiological time-series analysis using approximate entropy and sample entropy. Am J Physiol Heart Circ Physiol 2000;278: H2039–49. [9] Costa M, Goldberger AL, Peng CK. Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett 2002;89:068102.
586
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
[10] Wei Q, Liu DH, Wang KH, Liu Q, Abbod MF, Jiang BC, et al. Multivariate multiscale entropy applied to center of pressure signal analysis: an effect of vibration stimulation of shoes. Entropy 2012;14:2157–72. [11] Kim J, Andre´ E. Emotional recognition based on physiological changes in music listening. IEEE Trans Pattern Anal Mach Intell 2008;30(12):2067–83. [12] Mclntosh AR, Kovacevic N, Itier RJ. Increased brain signal variability accompany lower behavioral variability in development. PLoS Comput Biol 2008;4(7):e1000106. [13] Bosl W, Tierney A, Tager-Flusberg H, Nelson C. EEG complexity as a biomarker for autism spectrum disorder risk. BMC Med 2011;9(1):18.
[14] Rama-Lopez J, Pe´rez N, Martınez VE. Dynamic posture assessment in patients with peripheral vestibulopathy. Acta Otolaryngol 2004;124. 700A´/705. [15] Yeh JR, Hsu LC, Lin C, Chang FL, Lo MT. Nonlinear analysis of sensory organization test for subjects with unilateral vestibular dysfunction. PLoS ONE 2014;9(3):e91230. [16] Horak FB, Shupert CL, Mirka A. Components of postural dyscontrol in the elderly: a review. Neurobiol Aging 1989;10:727–38. [17] Manchester D, Woollacott M, Zederbauer-Hylton N, Marin O. Visual, vestibular and somatosensory contributions to balance control in the older adult. J Gerontol 1989;44:M118–27.
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Complexity of human postural control in subjects with unilateral peripheral vestibular hypofunction Jia-Rong Yeh a,b, Men-Tzung Lo a,b, Fu-Ling Chang c,d, Li-Chi Hsu e,f,* a
Research Center for Adaptive Data Analysis, National Central University, Taoyuan, Taiwan Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taoyuan, Taiwan c Taiwan Textile Research Institute, New Taipei City, Taiwan d National Yunlin University of Science and Technology, Graduate School of Design, Yunlin, Taiwan e Department of Neurology, Taipei Veterans General Hospital, Taipei, Taiwan f National Yang-Ming University School of Medicine, Taipei, Taiwan b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 20 January 2014 Received in revised form 16 April 2014 Accepted 30 June 2014
Complexity is a new measure for identifying the adaptability of a complex system to meet possible challenges. For a center of pressure (COP) time series, the complexity measure represents the stability of postural control. In this study, multiscale entropy (MSE) was used to evaluate the complexity of COP time series in six test conditions of sensory organization test (SOT). Complexity index (CI) is defined as the summation of entropies with coarse-graining scales 1–20 by MSE. A total of 51 subjects belonging to 3 groups – healthy-young, healthy-elderly and dizzy – were recruited in this study. The COP signals in both anteroposterior (AP) and mediolateral (ML) directions were analyzed respectively. According to our results, the CI of AP-direction COP time series is significantly correlated to the equilibrium score, which represents the stability of postural control in SOT. The AP-direction sway is significant larger than the ML-direction sway, particularly in the test conditions with sway-surface. In additions, the CI of APdirection COP for the healthy-elderly and dizzy groups are significantly lower than those for the healthy young group in the test conditions 1–4. The CI of ML-direction COP for the healthy-elderly group is significantly lower than those for the healthy-young and dizzy groups under test conditions 3 and 6. These results show that the complexity loss is a common status of AP-direction COP time series for both healthy-elderly and dizzy groups, and the complexity of ML-direction COP time series for subjects with unilateral vestibular dysfunction is higher than that for the healthy-elderly group specifically under test conditions 3 and 6. ß 2014 Elsevier B.V. All rights reserved.
Keywords: Complexity Center of pressure Multiscale entropy Sensory organization test Unilateral vestibular dysfunction
1. Introduction It is well-known that the integration of sensory systems, especially the visual, somatosensory and vestibular systems, are in close relation to the patients’ postural control [1]. In addition to direct observation of postural stability, posturography is a wellknown method to study balance disorders. The sensory organization test (SOT) paradigm of the posturographic test is a common protocol used in clinical practice. The principle of SOT evaluation is to selectively disrupt the support surface and/or the visual input to
* Corresponding author at: Department of Neurology, Taipei Veterans General Hospital, No. 201, Sec. 2, Shih-Pai Rd., Taipei 11217, Taiwan. Tel.: +886 2 28757578. E-mail addresses: [email protected], [email protected] (L.-C. Hsu). http://dx.doi.org/10.1016/j.gaitpost.2014.06.016 0966-6362/ß 2014 Elsevier B.V. All rights reserved.
measure the subject’s ability to use the remaining sensory inputs for postural control [2]. As the degree of task difficulty increased from open-eyes to closed-eyes and support surface from rigidsurface to sway-referencing surface, the instability of posture increased as well [3]. When analyzed properly, the SOT can determine which sensory systems (visual, somatosensory or vestibular) the patient relies upon most to maintain stability. In clinical practice, a formal SOT report includes the composite equilibrium scores in six test conditions and four ratios of sensory analysis. The four ratios are considered to reflect the functional conditions of somatosensory (SOM), visual (VIS), vestibular (VEST) and visual preference (PREF). In addition to the classic SOT report, the trajectory of the center of pressure (COP) time series can be measured in both anteroposterior (AP) and mediolateral (ML) directions respectively for all trials of SOT. In this study, COP time series in both AP and ML directions in six test conditions were
582
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
analyzed for characterizing the dynamical features of postural sway with disrupted and/or impaired sensory inputs in balance disorder patients. Aging and sensory disorders are important factors that affect the ability of postural control in human. The decrease of complexity of postural control system was proposed as a key signature of aging and sensory disorders [4–6]. In previous studies of complexity analysis, various algorithms have been proposed to estimate information entropy, i.e., complexity [7–10]. Multiscale entropy (MSE) analysis is based on the definition of the sample entropy (SampEn) proposed by Richman and Moorman [8]. Costa et al. defined a new measurement of ‘complexity index (CI)’ as the summation of the values of SampEn on multiple time scales [9]. The CI of both ML- and AP-direction COP time series represents the dynamical features of responses of postural control. So far, the CI acts as a good measurement of complexity in many applications [11–14]. In this study, the CI derived by the MSE method was used to evaluate the complexity features of COP time series in different test conditions of SOT for a total of 51 subjects with differences in age and functional conditions of vestibular systems. According to our results, the CI of COP time series is significantly correlated with the stability of postural control. A significant correlation between CI and equilibrium score (EQS) is identified by Pearson’s correlation coefficient with a value of 0.70. Moreover, the complexity of AP sway is negatively correlated to age in test conditions 1–4. The difference between the complexity of APdirection COP time series for the healthy elderly (HE) and dizzy (DZ) group is not significant. On the other hand, complexity of MLdirection COP time series represents as a critical assessment for identifying the dynamic property of postural control for the DZ group in test conditions 3 and 6, which shows significantly differences in comparisons with that of the HE group. 2. Methodology 2.1. Subjects The study was conducted at the Taipei Veterans’ General Hospital, which is a tertiary referral hospital in northern Taipei, Taiwan. The 51 subjects were divided into three groups: a healthy young (HY) group (19 males, 4 females, mean age: 36.2 5.7, range: 24–48), a healthy elderly (HE) group (3 males, 6 females, mean age: 68.6 5.2, range: 62–75) and a dizzy/imbalance (DZ) group (10 males, 9 females, mean age: 66.7 13.4, range: 40–85). Healthy subjects (HY and HE) had no past history of peripheral or central vestibular disorders, other medical or neurological diseases known to cause dizziness or gait disturbance and normal vestibular function as assessed by classical clinical otoneurologic tests including videooculographic and caloric tests. DZ patients all complained of chronic (>1 month) dizziness and unsteadiness. In addition, they all had unilateral peripheral vestibular hypofunction as documented by the caloric test (unilateral canal paresis 325%). Otherwise, none of the patients had any central lesions that could be interpreted as the origin of their dizziness or disequilibrium after appropriate clinical and/or neuroradiological examinations. The study hospital’s Institutional Review Board approved this trial and informed consent was obtained from each participant. 2.2. Sensory organization test (SOT) Computerized dynamic posturography is a functional test of the visual, vestibular and proprioceptive systems’ contribution to the maintenance of upright posture. The testing provides patients with a combination of present, absent, or distorted visual or proprioceptive cues based on a movable platform support and visual
surrounding. In the present study, the SMART Balance Master (NeuroCom, Clakamas, OR, USA) platform posturography was used for evaluation. The SOT protocol objectively identifies abnormalities in the patient’s use of the three sensory systems that contribute to postural control: somatosensory, visual and vestibular. Six test conditions for the SOT were designed: (1) eyes open with fixed surrounding and support; (2) eyes closed with fixed support; (3) sway-referenced surrounding with fixed support; (4) eyes open with fixed surrounding and sway-referenced support; (5) eyes closed with sway-referenced support; (6) sway-referenced surroundings and support. Equilibrium score (EQS) were measured as a nondimensional percentage comparing individual’s peak amplitude of AP sway with limit of stability (LOS). The sensory analysis aims to interpret the functional compensations in sensory systems using ratios of EQS in different test conditions. Each subject undertook 3 successful trials to keep standing without falling under 6 test conditions. If a subject fell in test, he was asked to take a break and one more trial. The COP time series was derived from the raw recordings of the force platform at a sampling rate of 100 Hz. 2.3. Multiscale entropy (MSE) The MSE method aims to measure the degree of complexity of a time series on multiple time scales. CI is defined as the summation of entropies for coarse-grained time series with scales of 1 n. The entropy of a coarse-grained time series is evaluated by previously suggested method so-called sample entropy (SampEn) [8]. The basic idea of the definition of SampEn is to provide a measure of ‘predictability’ in a time series by testing if there are any repeated patterns of various lengths. To calculate the predictability in a time series with N data points, a sequence of m successive points is selected as a vector from all possible m-point vectors, i.e., a total of N m + 1. Then, all m-point vectors that are similar to the selected one are counted, excluding the selected on itself. The number of similar vectors is denoted as nr;m where the index i is denotes the i vector number and r is the similarity tolerance. A difference between two points smaller than similarity tolerance is denoted as two points are similar to each other. All points in a vector are similar to the corresponding points in another vector, then two vectors are similar. The ratio nr;m =ðNm þ 1Þ represents the i probability of finding an m-point vector similar to the selected one. The average of all probabilities, which is computed for every possible m-point vector in the time series, represents the probability that two randomly selected vectors are similar to each other. The successive point of an m-point vector from the time series added to an m-point vector to result a (m + 1)-point vector. Considering two vectors that are similar to each other, the two resulting (m + 1)-point vectors may or may not be similar to each other. As if the resulting (m + 1)-point vectors that are similar to each other, we then say the (m + 1)th point is predictable when the two m-point vector are similar to each other. Therefore, the predictability, denoted as Cr,m, can be defined as: C
r;m
¼
PNm
r;mþ1 i¼1 ni P Nmþ1 r;m 1 ni i¼1 ðNmþ1Þ 1 ðNmÞ
(1)
Theoretically, predictability represents the regularity and entropy represents the irregularity of a time series. Therefore, sample entropy (SampEn) can be defined as the natural logarithm of the inverse of Cr,m. This definition of SampEn was extended by Costa et al. [9] to define the MSE as a result of successive coarsegraining of the time series. The coarse-graining was done by averaging the data points in given non-overlapping windows, t is
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
the number of data length of window, called the coarse-graining scale. The entropies versus their coarse-graining scales can be shown as a curve in a MSE plot. The under-curve area represents the summations of the entropies on multiple coarse-graining scales as a new assessment denoted as complexity index (CI). 2.4. Statistical analysis The population distributions of the EQS and CI were more Poisson-like than Gaussian-like. Therefore, the nonparametric Kolmogorov–Smirnov test (K–S test) was used to verify the difference between the two populations. The level of significance for the K–S test was set at P < 0.05. 3. Results 3.1. Results of sensory organization test A classical routine SOT report contains EQS in six test conditions and four ratios of sensory analysis. Table 1 shows the results of SOT report. The age of the HY groups is significantly lower than the other two groups. The difference between the HE and DZ groups is insignificant. As shown in Table 1, the EQS of test 1 in HE group was significantly lower than those for the other two groups which mean poorer balance stability during quiet stance in this group. Moreover, the EQS of tests 4 and 5 in HY group were significantly higher than those of the other two groups. These imply that the balance stability of the HY group is better than the other groups. Based on subjects’ ages and vestibular function status, aging might be a contributing factor of the difference between the HY and HE groups and the difference between the HE and DZ group might be attributed to the dysfunction of vestibular system. 3.2. Results of complexity analysis by multiscale entropy (MSE)
[(Fig._1)TD$IG]
The COP time series in both AP and ML directions were analyzed by MSE method respectively. The MSE plots for the three groups in six test conditions are shown in Figs. 1 and 2. Fig. 1 shows the MSE plots of ML-direction COP time series and Fig. 2 shows those for AP-direction COP time series. The CI was defined as the summation of entropies on the first 20 coarse-graining scales in this study. Therefore, the area under an MSE plot represents the complexity of a COP time series in a specific test condition. As shown in Figs. 1 and 2, the complexity of both ML- and AP-direction COP time series for the HY group is the highest one among three groups in all test conditions. In test condition 1, the subjects were instructed to stand on the fixed force plate with eyes opened. No sensory inputs of postural control are disrupted in test condition 1, so the complexity of COP time series is the highest. Whereas anyone of the sensory input is disrupted in the other test conditions, the complexity of COP time series is declined. On the other
583
Table 1 Statistical results of the original SOT reports comprising equilibrium scores (EQS) of six test conditions and four ratios of sensory analysis. The values were shown in mean standard deviation. Parameter
Healthy young
Dizzy
Healthy elderly
Subject number Age EQS of C1 EQS of C2 EQS of C3 EQS of C4 EQS of C5 EQS of C6 SOM VIS VEST PREF
23 36.2 5.7dd,ee 95.67 1.15ee 92.06 1.96 91.87 3.89e 87.28 4.58d,e 67.65 9.03d,ee 70.39 13.80d .962 .017 .912 .047d .707 .093dd,e 1.018 .086
19 68.6 5.2yy 95.44 1.35ee 92.18 3.01 89.33 4.83 82.07 6.35y 56.56 8.42y 55.07 15.23y .966 .026 .860 .060y .593 .089yy .972 .100
9 66.7 13.4yy 93.33 1.59yy,dd 90.54 2.16 87.38 5.90y 82.08 6.77y 56.25 14.32yy 62.96 9.74 .970 .017 .879 .064 .601 .145y 1.027 .048
Mark of ‘y’ means statistically significant difference in comparing with healthy young. Mark of ‘d’ means statistically significant difference in comparing with dizzy. Mark of ‘e’ means statistically significant difference in comparing with healthy elderly. Single character means with P-value < 0.05; twin characters means P-value < 0.01. hand, the complexities of AP-direction COP time series are significantly lower than those of ML-direction COP time series by comparing the results shown in Figs. 1 and 2. Fig. 3 depicts the results of CI for three groups in SOT. As shown in Fig. 3, the declines of complexity are strongly correlated to the difficulties of postural control in various test conditions. Sway-referenced surface causes the most significant complexity decline in comparing the CI values in test conditions of 1 and 4. The increasing instabilities of postural control caused by different sensory disruptions can be also double checked in sensory analysis. In addition, a significant correlation between EQS and the CI of AP-direction COP time series is verified by Pearson’s correlation coefficient with value of 0.6995. The correlation coefficient between EQS and CI of ML-direction COP time series is 0.6620. Therefore, CI might be considered a good assessment tool for the evaluation of balance stability. Furthermore, the complexity features of ML-direction COP time series are significantly different from those of AP-direction COP time series. The CI of APdirection COP time series in the HY group are significantly higher than those of the HE and DZ groups in test conditions 1–4. These results imply that aging is responsible for the reduction of complexity in AP-direction COP time series. The differences between the CI of AP-direction COP time series of HE and DZ groups in all of the six conditions of SOT are insignificant. In summary, the complexity decline of AP-direction time series is significant for elderly subjects regardless of the presence or absence of unilateral vestibular dysfunction.
Fig. 1. The MSE plots for the ML direction COP time series under six SOT test conditions.
[(Fig._2)TD$IG]
584
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
Fig. 2. The MSE plots for the AP direction COP time series under six SOT test conditions.
On the other hand, the CI values of ML-direction COP time series represent different dynamical features in comparison with those of AP-direction COP time series. The CI values of ML-direction COP time series for the HY group is the highest one and those for the HE group is the lowest one among the three groups. These results delineate that the complexity decline in the ML-direction COP time series for the DZ group is less than that for the HE group. Of note, there are significant differences between the CI values of ML-direction COP time series for the HE and DZ groups in test conditions 3 and 6. This finding implies that the CI of ML-direction COP time series in test conditions 3 and 6 might potentially reflect the specific features of posturographic evaluation for subjects with unilateral vestibular pathology.
4. Discussion
[(Fig._3)TD$IG]
In this study, the CI calculated by the MSE method was used to be a new measurement for assessing the postural control mechanism. As a general hypothesis, a system with high complexity reflects high capability to adapt to a constantly changing environment. Therefore, a higher CI value represents a better ability to maintain postural stability. Both EQS and CI are good tools for the assessment of balance ability of postural control. However,
the EQS is measured as the ratio between individual’s peak amplitude of AP sway and limit of stability, which represents a linear parameter directly associated with the magnitude of AP sway. Different from the linear parameter of EQS, the CI is a nonlinear parameter associated with the dynamical property of COP time series, which is independent from the magnitude of postural sway. A decrease of balance stability may cause an increase of postural sway magnitude and reduce the complexity of COP time series simultaneously. The changes of sway magnitude are essentially different from the changes of complexity of COP time series. Therefore, the performances of these two parameters should be different in posturography analysis and the analysis results should be compensative to each other. In this investigation, we have proved that the complexity of COP time series in specific test conditions of SOT actually reflects the differences in posturographic performance between the groups with and without unilateral vestibular dysfunction, which cannot be verified by the parameters based on linear assessment of EQS.
Fig. 3. The statistical results of the complexity index (CI) for three groups under six test conditions of SOT. The data is shown in mean values and standard errors.
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586 Table 2 Statistical results of complexity index (CI) values for AP- and ML-direction CoP time series. The values were shown in mean standard deviation. Healthy young For ML-direction of CoP time series CI in test 1 19.87 4.49e CI in test 2 20.46 4.55dd,ee CI in test 3 19.04 4.87e CI in test 4 18.87 3.62e CI in test 5 17.57 4.42d,e CI in test 6 17.83 4.44ee For AP-direction of CoP time series CI in test 1 18.07 4.05dd,ee CI in test 2 18.03 4.48d,ee CI in test 3 17.02 4.80d,ee CI in test 4 17.81 4.57dd,ee CI in test 5 13.97 5.49 CI in test 6 15.23 5.89e
Dizzy
Healthy elderly
19.18 4.48 18.14 5.68yy 18.74 4.86ee 17.98 4.54 16.14 4.58y 16.86 4.61e
17.67 3.59y 15.53 3.88yy 16.31 4.48y,dd 16.20 4.69y 15.30 3.76y 14.20 3.87yy,d
15.93 5.00yy 15.42 5.34y 13.87 5.25y 13.97 5.54yy 13.41 5.66 13.86 6.20
15.34 3.88yy 15.21 4.67yy 13.86 4.46yy 14.06 4.33yy 13.42 4.27 12.59 4.20y
Mark of ‘y’ means statistically significant difference in comparing with healthy young. Mark of ‘d’ means statistically significant difference in comparing with dizzy. Mark of ‘e’ means statistically significant difference in comparing with healthy elderly. Single character means with P-value < 0.05; twin characters means P-value < 0.01.
Moreover, most of the subjects with unilateral vestibular dysfunction are elderly in this study. Therefore, we also recruited healthy elderly subjects with age distribution similar to that of the subjects with unilateral vestibular dysfunction. So, the difference between the HE and DZ groups might represent the difference caused by the vestibular dysfunction. As shown in Table 1, only the EQS in test condition 1 revealed significant difference between HE and DZ groups. It is interesting that the EQS in test condition 1 for the HY group is insignificantly different from that for the DZ group. It implies that subjects with chromic unilateral peripheral vestibular dysfunction might have good compensation of postural controls in stress-free stance based on the results of EQS. The compensatory effect of the normal side of inner ear might keep the stability of postural control in the test condition of stress-free stance. On the other hand, the CI of AP-direction COP time series also does not differentiate those changes due to aging process from those caused by the unilateral peripheral vestibular dysfunction as shown in Table 2. The CI values of AP-direction COP time series show that aging causes the decline of complexity in AP-direction sway in test conditions 1–4. The complexity of AP-direction COP time series for the HY group is significantly higher than those for HE and DZ groups. On the contrary, the CI value of ML-direction COP time series presents different results in classifications among these three groups. In test conditions 3 and 6, the CI value of MLdirection COP time series for the DZ group is significantly higher than that for the HE groups. This result implies that subjects with unilateral vestibular dysfunction had better balance ability for maintaining ML-direction sway than healthy elderly subjects in the test conditions with sway-referenced visual surrounding. As the report in a previous study, the effects of visual stimulation on balance, such as test conditions 3 and 6, in patients with unilateral vestibulopathy reflect the type of sensory deficit and can be considered to be specific to such a deficit [14]. According to our observations reported in a previous study [15], sway referencing in the AP direction caused increased low-frequency fluctuations of AP-direction COP time series for both HE and DZ groups. However, changes of low-frequency fluctuations in ML-direction were insignificant for the DZ group. This implies that sway referencing in AP direction caused the increased low-frequency fluctuations in both directions of AP and ML for the HE group, but only in AP direction for the DZ group. The increased low-frequency fluctuation can cause the increased standard deviation of COP time series
585
and the decreased CI in a complexity analysis by MSE. Thus, CI in AP direction for the DZ group is similar to that for the HE group, and CI in ML direction for the DZ group is significantly higher than that for the HE group. In addition, visual input was disturbed by eye closing and both somatosensory and vestibular inputs were disturbed by sway supporting in AP direction for the test condition of SOT 5. There are no significant differences in sensory inputs among three groups for test condition of SOT 5. It resulted in similar AP-direction MSE plots and CI values for three groups in test condition of SOT 5. Moreover, why the elderly subjects performed even less than the DZ group? Decline in the integrity and functionality of sensory systems associated with advancing age [16,17] might play an important role because this may lead to postural control deficit in the elderly. Further studies are needed to test the role of conflict visual input in the maintenance of postural stability in the elderly. Nonetheless, our finding implies that the compensation of unilateral vestibular input drove a significant increase of ML-direction complexity as a specific pattern in posturography. In summary, the technique of complexity analysis was used to investigate the dynamic properties of COP time series in different test conditions of SOT. The nonlinear parameters derived by complexity analysis present different classification results in comparison with linear parameters. We suggest that the nonlinear complexity analysis is valuable for investigating the properties of a dynamic system with multiple control mechanisms. A welldesigned test protocol and an appropriate analysis algorithm will benefit to detail the functional conditions of a complicated system, such as the postural control system. Acknowledgements The authors wish to express their thanks for supports from the National Science Council (Taiwan) (Grant No 99-2627-B-008-003, and 97-2314-B-038-002-MY3), for joint funding from National Central University (Grant Nos CNJRF-99CGH-NCU-A3 and VGHUST100-G1-4-3), NSC support for the Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan (NSC 99-2911-I-008-100); and grants from the Taipei Veterans General Hospital (V99C1-194, V100C-205 and V101C-200). Conflict of interest statement. The authors disclose there are no potential conflicts of interest with other people or organizations in this study.
References [1] Woolacott MH, Shumway-Cook A. Attention and the control of posture and gait: a review of an emerging area of research. Gait Posture 2002;16:1–14. [2] Nashner LM. Practical management for the dizzy patient: computerized dynamic posturography. Philadelphia, PA: Lippincott Williams & Wilkins; 2008. [3] Negahban H, Salavati M, Mazaheri M, Sanjari AM, Hadian MR, Parnianpour M. Non-linear dynamical features of center of pressure extracted by recurrence quantification analysis in people with unilateral anterior cruciate ligament injury. Gait Posture 2010;31:450–5. [4] Goldberger AL, Amaral LA, Hausdorff JM, Ivanov P, Peng CK, Stanley HE. Fractal dynamics in physiology: alternations with disease and aging. Proc Natl Acad Sci USA 2002;99:2466–71. [5] Goldberger AL, Peng CK, Lipsitz LA. What is physiologic complexity and how does it change with aging and disease? Neurobiol Aging 2002;23:23–6. [6] Lipsitz LA. Dynamics of stability: the physiologic basis of functional health and frailty. J Gerontol A: Biol Sci Med Sci 2002;57:B115–25. [7] Pincus SM. Approximate entropy as a measure of system complexity. Proc Natl Acad Sci USA 1991;88:2297–301. [8] Richman JS, Moorman JR. Physiological time-series analysis using approximate entropy and sample entropy. Am J Physiol Heart Circ Physiol 2000;278: H2039–49. [9] Costa M, Goldberger AL, Peng CK. Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett 2002;89:068102.
586
J.-R. Yeh et al. / Gait & Posture 40 (2014) 581–586
[10] Wei Q, Liu DH, Wang KH, Liu Q, Abbod MF, Jiang BC, et al. Multivariate multiscale entropy applied to center of pressure signal analysis: an effect of vibration stimulation of shoes. Entropy 2012;14:2157–72. [11] Kim J, Andre´ E. Emotional recognition based on physiological changes in music listening. IEEE Trans Pattern Anal Mach Intell 2008;30(12):2067–83. [12] Mclntosh AR, Kovacevic N, Itier RJ. Increased brain signal variability accompany lower behavioral variability in development. PLoS Comput Biol 2008;4(7):e1000106. [13] Bosl W, Tierney A, Tager-Flusberg H, Nelson C. EEG complexity as a biomarker for autism spectrum disorder risk. BMC Med 2011;9(1):18.
[14] Rama-Lopez J, Pe´rez N, Martınez VE. Dynamic posture assessment in patients with peripheral vestibulopathy. Acta Otolaryngol 2004;124. 700A´/705. [15] Yeh JR, Hsu LC, Lin C, Chang FL, Lo MT. Nonlinear analysis of sensory organization test for subjects with unilateral vestibular dysfunction. PLoS ONE 2014;9(3):e91230. [16] Horak FB, Shupert CL, Mirka A. Components of postural dyscontrol in the elderly: a review. Neurobiol Aging 1989;10:727–38. [17] Manchester D, Woollacott M, Zederbauer-Hylton N, Marin O. Visual, vestibular and somatosensory contributions to balance control in the older adult. J Gerontol 1989;44:M118–27.
