This article was downloaded by: [Eindhoven Technical University] On: 13 February 2015, At: 23:25 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Motor Behavior Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vjmb20
Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss a
a
Fellipe M. Portela & Arthur S. Ferreira a
Laboratory of Computational Simulation and Modeling in Rehabilitation, Centro Universitário Augusto Motta, Rio de Janeiro, Brazil Published online: 19 Jun 2014.
Click for updates To cite this article: Fellipe M. Portela & Arthur S. Ferreira (2014) Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss, Journal of Motor Behavior, 46:5, 369-379, DOI: 10.1080/00222895.2014.916651 To link to this article: http://dx.doi.org/10.1080/00222895.2014.916651
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Journal of Motor Behavior, Vol. 46, No. 5, 2014 Copyright © Taylor & Francis Group, LLC
RESEARCH ARTICLE
Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss Fellipe M. Portela, Arthur S. Ferreira
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
rio Augusto Motta, Laboratory of Computational Simulation and Modeling in Rehabilitation, Centro Universita Rio de Janeiro, Brazil. ABSTRACT. The spatial distribution of center-of-pressure speed during postural tasks and its changes due to somatosensory constraint (temporary ischemic hypoxia on ankle/feet) were investigated in young, healthy subjects (n D 13). A single high-speed region in the central region of the statokinesigram was observed during postural tasks with full sensory information. A significant increase in the quantity of high-speed regions was observed during ischemia and somatosensory constraint, whereas a significant increase in the quantity of high-speed regions localized more distant to the center of center-of-pressure area occurred under somatosensory constraints, suggesting a redirection of center-ofpressure trajectory to adjust the position of the center of mass with respect to the egocentric reference of balance.
Lin, 2008), and muscle fatigue (Hlavackova & Vuillerme, 2012). Because the combination of CoP position and velocity in a phase-plane plot was regarded as an important tool for assessment of postural control (Bottaro, Yasutake, Nomura, Casadio, & Morasso, 2008; Riley, Benda, GillBody, & Krebs, 1995) and was recently argued as a predictor for risk of fall (Honarvar & Nakashima, 2013) both in an undisturbed upright stance, it is possible that other methods combining kinematic variables provide new insights about the postural control. Despite the hypothesis that a large CoP area may be a natural exploratory behavior for postural control and not a consequence of postural instability (Carpenter, Murnaghan, & Inglis, 2010), higher CoP area and velocity are usually interpreted as worse postural control (Raymakers, Samson, & Verhaar, 2005). Considering the relationships between position and velocity as kinematic variables, an interesting question arises: Where is it necessary to use high CoP velocities for maintaining the upright posture? On the one hand, high CoP velocities should be avoided near the boundaries of the base of support since there may be not enough time for an efficient postural adjustment at these locations (Riccio, 1993). Indeed, the velocity-based postural control was described to occur in the central region of the stabilogram, where CoP velocity reaches its maximal absolute values (Delignieres et al., 2011a), although no evidence was presented regarding this observation. On the other hand, high CoP velocities near the boundaries of the base of support may be necessary to quickly redirect the body’s center of mass toward the egocentric reference of posture in more challenging conditions or after a fall is initiated. Therefore, a method to assess the spatial distribution of regions with high CoP velocity would be valuable for investigation what postural strategies do patients at high risk of falling use. Presently, the spatial distribution of regions with high CoP velocity remains uninvestigated. This work describes the spatial distribution of CoP velocity during the undisturbed upright stance in healthy subjects; a companion method is also introduced for visualization, localization, and quantification of regions with high CoP velocity, namely kinematic mapping. Explicitly stating, this study was motivated by the velocity-
Keywords: center-of-pressure, egocentric constraints, kinematics, postural sway, rehabilitation
D
isplacements of the body’s center of mass while subjects perform postural tasks are described by anteroposterior (AP) and medialateral (ML) coordinates of the body’s center of pressure (CoP; Duarte & Freitas, 2010; Winter, 1995). From the time-sampled CoP coordinates, it is possible to calculate statistical, geometrical, kinematic, structural, and fractal variables related to postural control in one or two spatial dimensions (Baratto, Morasso, Re, & Spada, 2002; Deschamps, Beauchet, Annweiler, Cornu, & Mignardot, 2014; Duarte & Freitas, 2010; Kapteyn et al., 1983). Previous studies (Collins & De Luca, 1993; Duarte & Zatsiorsky, 2002; Juras, Stomba, Fredyk, Sobota, & Bacik, 2008; Zatsiorsky & King, 1998) investigated the postural control in the undisturbed upright stance considering parameters calculated from CoP displacement data. More recent studies emphasized the role of velocity for postural control (Jeka, Kiemel, Creath, Horak, & Peterka, 2004; Masani, Popovic, Nakazawa, Kouzaki, & Nozaki, 2003; Pai & Patton, 1997; Sasagawa, Ushiyamab, Kouzakic, & Kanehisa, 2009) and provided support to the velocity-based theory of an intermittent control of the upright posture (Delignieres, Torre, & Bernard, 2011a). Particularly, kinematic variables (e.g. linear path, displacement, and velocity) are of clinical interest because they reflect postural responses due to ageing (Du Pasquier et al., 2003), chronic ankle instability (Wikstrom, Fournier, & McKeon, 2010), Parkinson’s disease (Mitchell, Collins, De Luca, Burrows, & Lipsitz, 1995), diabetic peripheral neuropathy (Boucher, Teasdale, Courtemanche, Bard, & Fleury, 1995), lower limb somatosensory loss (Wang &
Correspondence address: Arthur de S a Ferreira, Pra¸c a das Na¸c o~es 34, 3 andar, Bonsucesso, Rio de Janeiro, RJ – Brazil, CEP 21041–010. e-mail: [email protected] 369
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
based theory for controlling the undisturbed upright posture, which hypothesizes that CoP velocity is bounded and body sway is left uncheck until a threshold in velocity is reached (Delignieres et al., 2011a). In this theory, the CoP signal and related parameters are used because they reflect the overall process underlying the postural control scheme (Delignieres et al., 2011a; Winter, 2005). The clinical utility of this method regarding identification of spatial patterns under distinct postural challenges was verified with a model of somatosensory loss temporarily induced by an ischemic hypoxia of the lower limbs. It was hypothesized that during unchallenging postural tasks subjects will exhibit one highspeed region (HSR) centered at the CoP area, and that the suppression of somatosensory input would increase both the quantity of HSR as well as the distance away to the center of the CoP area. Method Ethics Statement The Institutional Ethics Committee approved this study before its execution (CAAE-02420912.8.0000.5235). All volunteers were informed about the procedures and gave their written consent to participate in the study. Subjects Thirteen healthy volunteers (eight women; 27 § 5 years old; 69.3 § 8.9 kg; 1.68 § 0.09 m tall) from the academic community agreed to participate in this study. Subjects were admitted after passing both clinical screening and physical examination to exclude musculoskeletal, neurological, and vascular disorders that might compromise their safety and the execution of the protocol. This study was conducted in a silent room with controlled air temperature (21–23 C). Blocks of Postural Tasks Subjects were instructed to step up on the platform barefooted and to keep an undisturbed bipedal position with their arms hanging at their body sides (Figure 1, top panel). They performed a first block lasting 60 s that was preceded by a 20-s period to avoid recording movement artifacts due to stepping up. Combinations of visual and biomechanical constraints characterized the two postural tasks investigated: feet apart (8 cm between heels, 10 angle) with eyes opened (FAEO) and feet together (heels in contact, 10 angle) with eyes closed (FTEC). During tasks with eyes opened, volunteers were instructed to focus on a small target fixed at the eye level on a wall 1.5 m apart from the force platform. Between postural tasks volunteers were allowed to rest during 60 s. These two tasks were performed without wearing the pressure cuffs because they could have augmented the cutaneous sensory information at that level, thus influencing the postural control (Wang & Ling, 2008). 370
FIGURE 1. Positioning for postural tasks and somatosensory constraint. Top: Lower extremity positioning (left: feet apart; right: feet together) before ischemic hypoxia. Bottom: Lower extremity positioning (left: feet apart; right: feet together) during ischemic hypoxia with the sphygmomanometer cuff.
A second block was performed with feet and ankle ischemic hypoxia to generate a temporary somatosensory loss on these body parts (Wang & Lin, 2008). Blood pressure was measure in both arms and legs with an automated device (model G-TECH, BP3AF1-3, Onbo Electronics Co., Shenzen, China) to calculate the ankle-brachial index (group results: right side D 1.2 § 0.1, left side D 1.3 § 0.2) for ruling out subclinical peripheral vascular disease (Al-Qaisi, Nott, King, & Kaddoura, 2009). Two sphygmomanometer cuffs connected by a T tube were positioned above the lateral malleolus of the ankle joint (Figure 1, bottom panel) to interrupt the blood flow and consequently to impair conduction from deep and superficial mechanoreceptors above the level of the cuffs (Horak, Nashner, & Diener, 1990). The cuffs were manually inflated 75 mm Hg above the systolic blood pressure (higher measure from both arms) for 25 min and monitored regularly on both legs. Both postural tasks were repeated after 25 min of ischemic hypoxia with the cuffs still inflated. Following the last task, muscle strength was immediately assessed to evaluate the preservation of efferent innervation. Finally, both cuffs were deflated and the local reactive hyperemia was observed during 5–10 min before discharging the subject of the experiment.
Motor and Sensory Measurements Motor function test was performed based on standard manual testing (Hislop & Montgomery, 1996). Muscle groups assessed included the extensor digitorum longus and brevis, extensor hallucis longus, tibialis anterior, flexor hallucis brevis, lumbricales, and gastrocnemius. Journal of Motor Behavior
Kinematic Mapping of Center of Pressure
affirmative response was sufficient to characterize the respective level of touch-pressure threshold. Instrumentation, Signal Acquisition, and Preprocessing Figure 2 summarizes the computational methods proposed in this study for kinematic mapping of the HSR as implemented in LabVIEW 8.0 (National Instruments, Dallas, TX) running on Windows 7 (Microsoft Corporation, Seattle, WA). Signals were acquired at 100 Hz from analog channels of the force platform (AccuSwayPlus, AMTI, Watertown, MA) and digitalized by a 16-bit converter (§10 V, NI-USB 6210, National Instruments). Digital
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Verbal analog scale was applied every 2 min to assess the level of pain or discomfort from 0 (no discomfort or pain) to 10 (severe discomfort or pain). After 15 min of ischemic hypoxia, cutaneous sensitivity was monitored in a subgroup (n D 5) every 2 min by Semmes-Weinstein monofilament testing (SORRI-Bauru, S~ao Paulo, Brazil) according to the manufacturer’s instructions and previous studies (Wang & Lin, 2008). Briefly, monofilaments were applied in sequence (green: 0.05 g, blue: 0.2 g, violet: 2.0 g, red: 4.0 g, orange: 10.0 g, magenta: 300 g) with the subjects with eyes closed to eliminate visual cues. They were asked to answer “yes” whenever they felt the monofilament. The test was performed three times at the same place such that a single
FIGURE 2. Flow chart for signal acquisition, processing, and analysis of the statokinematogram. Computational methods proposed for generation, preprocessing, and analysis of the statokinematograms. AAMVx D average absolute maximal velocity on X-axis; AAMVy D average absolute maximal velocity on Y-axis; AAMVxy D average absolute maximal velocity on XY plane; CoP D center of pressure; Davg D average Euclidean distance from centroid; Dmax D maximal Euclidean distance from centroid; HSR D high-speed regions; SDVx D standard deviation on X-axis; SDVy D standard deviation on Y-axis; VX D Velocity on X-axis; VXmax D maximal velocity on X-axis; VXY D velocity on XY plane; VXYavg D average velocity on XY plane; VY D velocity on Y-axis; VYmax D maximal velocity on Y-axis.
2014, Vol. 46, No. 5
371
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
signals were stored for off-line processing and used to calculate forces (Fx, Fy, Fz) and moments of force (Mx, My, Mz) with the calibration matrix provided by the platform’s manufacturer as in a previous study (Baracat & de Sa Ferreira, 2013). CoP coordinates were calculated according to the manufacturer’s manual. CoP univariate signals were detrended and had their mean value subtracted before a low-pass digital filter (second order, Butterworth filter, cutoff frequency of 2.5 Hz) was applied in direct and reverse direction (Vieira, Oliveira, & Nadal, 2009). This cutoff frequency is justified simultaneously by the very low amplitude harmonics above 1 Hz in healthy subjects under undisturbed bipedal stance (Vieira et al., 2009) and the need of smooth signals for calculation of time derivatives of CoP displacement (Winter, 2005). CoP signals processed at this stage were used to compute the following variables related to body sway in both ML (X-axis) and AP (Y-axis) directions (Raymakers et al., 2005): standard deviation (SDX, SDY), and maximum velocity (VXmax, VYmax) from stabilograms; elliptical area using the principal component analysis method (Area; Oliveira, Simpson, & Nadal, 1996); and average velocity (Vavg) from statokinesigrams. Calculation of CoP Velocity and Speed CoP velocity was calculated using the finite difference calculus and the central difference method to each coordinate X and Y separately (Vx and Vy, respectively). CoP velocity in the ith sample (equations 1–2) was estimated considering the adjacent samples i–1 and iC1, thus providing a time alignment between the reference position for calculating the CoP velocity and the spatial coordinates of the XY axes (Robertson, Caldwell, Hamill, Kamen, & Whittlesey, 2004; Winter, 2005): xi C 1 ¡ xi ¡ 1 2Dt yi C 1 ¡ yi ¡ 1 Vyi D 2Dt Vxi D
(1) (2)
Where Dt represents the acquisition time interval, i corresponds to the sample index, xi–1 and xiC1 are the CoP coordinates at the X-axis in the samples surrounding the ith sample, and yi–1 and yiC1 are the CoP coordinates at the Yaxis in the samples surrounding the ith sample. The CoP speed in the XY plane (Vxy) for the ith sample was estimated using equation 3: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Vxyi D Vx2i C Vy2i
(3)
Kinematic Mapping of HSR The method of kinematic mapping of HSR is defined as the spatial localization of CoP speed in synchronization with the CoP bivariate time series. The aim of this method 372
is to map areas with CoP speed higher than an empirical threshold (Delignieres et al., 2011b). The thresholds are estimated as the average absolute maximal velocity (AAMV) from a 2-s epoch of Vx and Vy, namely AAMVx and AAMVy respectively. Likewise, a new threshold related to CoP bivariate signal was proposed, namely the average absolute maximal velocity of 2-s epochs of Vxy (AAMVxy). Therefore, regions in which the CoP spatial coordinates have Vxy greater than AAMVxy are denominated HSR. Because AAMVx and AAMVy thresholds are affected by postural tasks (Delignieres et al., 2011b) all thresholds were estimated for each postural task. The statokinematogram is a companion graph for visualization of HSR as obtained with the kinematic mapping. A zero-value bidimensional matrix (bin resolution: 1 £ 1 mm) was generated from the CoP bivariate series. A comparison is then performed between Vxy and AAMVxy: if the ith sample of Vxy estimated for the respective bin coordinate (Xi,Yi) is higher than AAMVxy, its value is summed to the previous value in the respective bin. At the end of this loop the cumulative sum in each bin is divided by the quantity of samples stored on the respective bin to obtain the CoP average speed at the respective bin. Since this map present low spatial resolution and quantization error due to categorization of continuous CoP velocity values into discrete bins, a median spatial filtering with a 3 £ 3 kernel size followed by spatial interpolation with bilinear algorithm (Pedrini & Schwartz, 2008) was performed (Figure 3). See also Movie 1, “Generation of Three-Dimensional Statokinematogram”, provided as an online supplementary file available at www.tandfonline.com/vjmb. Localization of HSR and Quantification of HSR in the Statokinematogram The bidimensional matrix with CoP average speed above AAMVxy was converted to a binary matrix with bins marked either as above (D1) or equal or below (D0) this threshold. In sequence, a two-scan connected-component labeling algorithm (He, Chao, & Suzuki, 2008) was applied to this binary matrix to locate HSRs. Because a variable quantity of HSR was expected, three variables were estimated using the centroid coordinates:
1. nHSR: the quantity of nonconnected HSR (i.e., HSR that do not share bins). This variable represents the quantity of distinct loci where CoP trajectory was changed for stabilization of posture. 2. Davg: the averaged Euclidean distance between all HSR centroids and the center of the CoP area. This variable represents the expected distances from HSR to the CoP area. 3. Dmax: the maximal Euclidean distance among all HSR centroids and the center of the CoP area. This variable represents the extreme distance from HSR to the CoP area, being the maximal value often regarded Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
FIGURE 3. Velocity time series in stabilograms and statokinesigrams, and high-speed regions on the statokinematogram. From top to bottom: Center-of-pressure (CoP) velocity time series from medialateral and anteroposterior stabilograms, and CoP speed time series from statokinesigram (all during feet apart–eyes opened, before ischemic hypoxia). Below is exhibited four statokinematograms from all combinations of postural tasks and somatosensory constraint from the same subject (#5). Dotted lines represent the thresholds for positive and negative axis and higher plane. Colored areas correspond to the high-speed regions (HSR) in the statokinematogram.
as a relevant factor when studying the biomechanics postural control (King & Zatsiorsky, 2002). Statistical Analysis Values in tables are presented as M § SD for continuous variables, median [minimum; maximum] for ordinal variables, and absolute frequency (%) for nominal ones. A twoway repeated-measures analysis of variance was performed to test the null hypothesis of no main or interaction effects related to postural task (levels: FAEO; FTEC) and somatosensory constraint (levels: no ischemia; ischemia) separately for variables of the stabilogram (SDX, SDY, VXmax, VYmax), statokinesigram (Vavg, Area), and statokinematogram (Dmax and Davg). Marginal homogeneity test was 2014, Vol. 46, No. 5
used to test the null hypothesis of equal marginal proportions of nHSR among all four levels. Association between variables was tested using the Spearman rank correlation coefficient (H0: r D 0) under FAEO with no ischemia. Statistical significance was considered at p < .05 with the adjusted p-value for multiple comparisons based on a stepwise rejection procedure (Li, 2008). Statistical analyses were performed in SPSS 17 (SPSS Inc., Chicago, IL). Robustness Analysis Methodological procedures affect the reliability of CoPbased parameters (Ruhe, Feier, & Walker, 2010). Therefore, a separated analysis was conducted to assess whether the kinematic mapping is robust to changes in signal processing 373
F. M. Portela & A. S. Ferreira
techniques. Two important methodological procedures used in this study match the recommendations for optimal reliability (trial duration 60 s and sampling frequency D 100 Hz). Thus is of particular interest the robustness analysis with respect to the low-pass filter cutoff frequency since in this study the selected cutoff frequency was lower than the recommended for processing CoP data (2.5 Hz instead of 10 Hz). The variables of the statokinematogram—nHSR, Davg, and Dmax—were recalculated using a 10 Hz lowpass filter for a subsequent statistical analysis using the same tests described previously for these variables.
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Results Sensory-motor analysis revealed that all subjects presented preserved muscle strength (grade 5) after 25 min of ischemia in every tested muscle. Also, all subjects reported numbness in the lower extremities and pain intensity averaged 4.4 § 1.6. Subjects tested with the monofilaments reported sensitivity at the violet-colored monofilament, interpreted as increased touch-pressure threshold. An overall significant difference in marginal proportions (p D .001) was observed among the four levels in nHSR (Table 1). Paired analysis of nHSR revealed significant marginal differences for somatosensory constraint during FTEC (p D .011), as well as for postural tasks during ischemia (p D .018). A significant interaction between postural task and somatosensory constraint was observed for Davg, Wilks’ L D .708, F(1, 12) D 4.961, p D .046, h2 D .292.
Dmax was significantly higher during FTEC, Wilks’ L D .485, F(1, 12) D 12.718, p D .004, h2 D .515; and somatosensory constraint, Wilks’ L D .586, F(1, 12) D 8.470, p D .013, h2 D .414, but without significant interaction (p D .053). AAMVx was significantly higher during FTEC, Wilks’ L D .211, F(1, 12) D 44.890, p < .001, h2 D .789, but not significantly different under somatosensory constraint, Wilks’ L D .831, F(1, 12) D 2.448, p D .144, h2 D .169. AAMVy was significantly higher during FTEC, Wilks’ L D .219, F(1, 12) D 42.778, p < .001, h2 D .781, and somatosensory constraint, Wilks’ L D .157, F(1, 12) D 64.521, p < .001, h2 D .843, but without significant interaction (p D .114). Finally, AAMVxy was significantly higher during FTEC, Wilks’ L D .207, F(1, 12) D 45.921, p < .001, h2 D .793, and somatosensory constraint, Wilks’ L D .195, F(1, 12) D 49.692, p < .001, h2 D .805, but without significant interaction (p D .141). Area was significantly higher during FTEC, Wilks’ L D .300, F(1, 12) D 27.967, p < .001, h2 D .700, and somatosensory constraint, Wilks’ L D .541, F(1, 12) D 10.166, p D .008, h2 D .459, but without significant interaction (p D .068). Vavg was significantly higher during FTEC, Wilks’ L D .173, F(1, 12) D 57.412, p < .001, h2 D .827, and somatosensory constraint, Wilks’ L D .299, F(1, 12) D 28.170, p < .001, h2 D .701, but without significant interaction (p D .190). SDx was significantly higher during FTEC, Wilks’ L D .170, F(1, 12) D 58.771, p < .001, h2 D .830, but not somatosensory constraint, Wilks’ L D .997, F(1, 12) D
TABLE 1. Comparative Analysis of Stabilometric Parameters According to the Postural Task and Somatosensory Constraints on Variables From the Stabilogram, Statokinesigram, and Statokinematogram Feet apart, eyes opened Variable Velocity thresholds AAMVx (cm/s) AAMVy (cm/s) AAMVxy (cm/s) Statokinematogram nHSR Davg (mm) Dmax (mm) Statokinesigram Area (cm2) Vavg (cm/s) Stabilogram SDX (cm) SDY (cm) VXmax (cm/s) VYmax (cm/s)
Feet together, eyes closed
Before hypoxia
During hypoxia
Before hypoxia
During hypoxia
p Values
4.20 § 0.79 7.12 § 1.55 19.43 § 3.44
4.22 § 0.72 8.49 § 2.07 23.22 § 5.64
8.98 § 2.98 12.42 § 4.01 35.62 § 10.46
9.68 § 3.48 15.31 § 4.78 43.17 § 13.37
< .001a; .144b < .001a; < .001b < .001a; < .001b
1 [1, 2] 1.67 § 1.56 2.21 § 2.41
1 [1, 4] 2.38 § 2.17 3.53 § 3.50
1 [1, 4] 3.04 § 2.26 4.69 § 4.12
3 [1, 5] 6.03 § 2.55 9.38 § 4.22
.001d .046d .004a; .013b
0.37 § 0.19 0.42 § 0.08
0.42 § 0.26 0.47 § 0.09
1.30 § 0.77 0.80 § 0.24
1.78 § 1.05 0.93 § 0.27
< .001a; .008b < .001a; < .001b
0.12 § 0.03 0.26 § 0.07 1.23 § 0.48 2.32 § 1.08
0.11 § 0.04 0.31 § 0.11 1.03 § 0.44 2.44 § 1.14
0.28 § 0.09 0.38 § 0.14 2.48 § 1.23 3.50 § 1.89
0.29 § 0.10 0.51 § 0.17 3.15 § 1.74 4.11 § 2.20
< .001a; .857b .001a; .001b .033c .018a; .408b
Note. Data format is M § SD or median [minimum, maximum]. a Main effect for postural task. bMain effect for somatosensory constraint. cInteraction effect between postural task and somatosensory constraint. d Friedman’s test.
374
Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
0.034, p D .857, h2 D .003. SDy was significantly higher during FTEC, Wilks’ L D .365, F(1, 12) D 20.870, p D .001, h2 D .635, and somatosensory constraint, Wilks’ L D .415, F(1, 12) D 16.887, p D .001, h2 D .585, but without significant interaction (p D .066). A significant interaction between postural task and somatosensory constraint was observed for VXmax, Wilks’ L D .674, F(1, 12) D 5.812, p D .033, h2 D .326. VYmax was significantly higher during FTEC, Wilks’ L D .614, F(1, 12) D 7.544, p D .018, h2 D .386, but not somatosensory constraint, Wilks’ L D .942, F(1, 12) D 0.736, p D .408, h2 D .058. Area was positively correlated to AAMVx (r D .569; p D .021), AAMVy (r D .664; p D .007), and AAMVxy (r D .641; p D .009). Variable nHSR was positively correlated to Davg (r D .804; p < .001) and Dmax (r D .804; p < .001). Conversely, Area was not significantly correlated to nHSR (r D .379; p D .101), Davg (r D .426; p D .073), or Dmax (r D .426; p D .073). Table 2 exhibits the results of the robustness analysis. Parameters nHSR, Davg, and Dmax were higher and generally more disperse if calculated with the 10 Hz cutoff frequency. Using this higher cutoff frequency yielded a similar behavior with a minor deterioration of the observed main and interaction effects. An overall significant difference in marginal proportions (p D .017) was maintained among the four levels in nHSR. Paired analysis of nHSR revealed again significant marginal differences for postural tasks during ischemia (p D .026) but not for somatosensory constraint during FTEC (p D .912). Davg was significantly higher during FTEC, Wilks’ L D .542, F(1, 12) D 10.144, p D .008, h2 D .458, but not under somatosensory constraint, Wilks’ L D .850, F(1, 12) D 2.121, p D .171, h2 D .150, without significant interaction (p D .774). Likewise, Dmax was significantly higher during FTEC, Wilks’ L D .500, F(1, 12) D 11.982, p D .005, h2 D .500, but not under
somatosensory constraint, Wilks’ L D .794, F(1, 12) D 3.109, p D .103, h2 D .206, again without significant interaction (p D .214). Discussion This work described the spatial distribution of CoP speed during the undisturbed upright stance in healthy subjects using the proposed method of kinematic mapping of HSR. The clinical utility of this method was verified with a model of somatosensory loss temporarily induced by ischemic hypoxia of the lower limbs, a method well-known for generating postural challenges in case of combined visual, biomechanical, and somatosensory constraints (Horak et al., 1990; Wang & Ling, 2008). The main results of this study comprise (a) the undisturbed upright stance is characterized by one HSR nearly centered at the CoP area, and (b) more challenging postural tasks increased both the quantity of HSR as well as their distance from the center of the CoP area. This is the first work to objectively localize and quantify HSRs, and most importantly, to show how their spatial distribution changes under somatosensory constraints. In other words, the observed results suggest that different velocity-based strategies for redirecting the body’s center of mass to the reference position area were used under more challenging postural tasks. Collectively, these results support that the kinematic mapping was sensitive to changes in postural task and somatosensory constraints. Most importantly, these results emphasize that the combination of CoP position and velocity provided by the kinematic mapping allows the assessment of the postural strategies used under different somatosensory and visual/ biomechanical constraints. Our hypothesis that a single HSR appears in the statokinematogram during postural tasks with full sensory
TABLE 2. Robustness Analysis of the Parameters From the Statokinematogram at Different Low-Pass Filter Cut-Off Frequencies Feet apart, eyes opened Variable nHSR 2.5 Hz 10.0 Hz Davg (mm) 2.5 Hz 10.0 Hz Dmax (mm) 2.5 Hz 10.0 Hz
Feet together, eyes closed
Before hypoxia
During hypoxia
Before hypoxia
During hypoxia
p Values
1 [1, 2] 2 [1, 4]
1 [1, 4] 1 [1, 4]
1 [1, 4] 3 [1, 7]
3 [1, 5] 3 [1, 9]
1.67 § 1.56 2.91 § 1.67
2.38 § 2.17 3.60 § 3.50
3.04 § 2.26 5.02 § 2.10
6.03 § 2.55 6.12 § 3.07
.046c .008a, .171b
2.21 § 2.41 4.41 § 2.90
3.53 § 3.50 4.62 § 4.04
4.69 § 4.12 7.19 § 3.56
9.38 § 4.22 9.73 § 4.92
.004a, .013b .005a, .103b
.001d .017d
Note. Data format is M § SD or median [minimum, maximum]. a Main effect for postural task. bMain effect for somatosensory constraint. cInteraction effect between postural task and somatosensory constraint. d Friedman’s test.
2014, Vol. 46, No. 5
375
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
information was confirmed. Although the velocity-based postural control has been suggested to occur in the central region of the stabilogram (CoP univariate analysis; Delignieres et al., 2011a), this is the first study to reveal the spatial localization of HSR in terms of CoP bivariate analysis. Also, our hypothesis of increased nHSR due to more challenging conditions was also confirmed for both postural task and somatosensory constraint during FTEC. Although the reason for occurrence of more HSR could not be determined in this study, we argued that such observation reflects the need for more changes in CoP trajectory redirecting the center of mass to maintain an upright posture under limited sensory input. Likewise, our hypothesis that suppression of sensory input increases the distance from HSR to the center of CoP area was confirmed. The interaction effect observed for Davg indicated that the effect of the absence of vision and change biomechanics was greater during ischemia for this variable. The fact that an increased quantity of HSR was localized more distant to the center of CoP area suggests that redirecting of CoP trajectory was more necessary to adjust the position of the center of mass with respect to the egocentric reference of balance. Although it may be argued that increased statokinematogram variables are expected due to also increased CoP area between postural tasks, the lack of significant correlations between Area and all statokinematogram variables do not favors this argument. Therefore, the observed changes is spatial distribution seems to be not explained by changes in CoP area and this fact needs to be further investigated. AAMVx showed a significant main effect for postural task, but not for somatosensory constraint, while AAMVy showed a significance main effect for both factors. This suggest that the AP threshold for redirecting the CoP was adjusted to fit the new condition of somatosensory constraint, which is in agreement with the sensory reweighting during limited sensory information (Redfern, Yardley, & Bronstein, 2001). Despite small differences between protocols and signal processing techniques, the calculated AAMVx and AAMVy in this study show similar behavior to those previously observed (Delignieres et al., 2011b) in relation to the group average values and the main effect under no visual input. However, no explanation was provided by the authors (Delignieres et al., 2011a, 2011b) to the observed increase in AAMVx and AAMVy as estimated with and without visual input. Considering the significant, positive correlation coefficients observed between Area and all three thresholds, our explanation is that a considerable amount of increase in CoP velocity threshold is explained by an increase in CoP area. The remaining increase in threshold may be interpreted as an error margin to tolerate higher CoP displacements under conditions of limited sensory input. Further studies should investigate if there is a causal relationship between CoP area and those empirical velocity thresholds, and if the increase in AAMV that is not explained by increase in Area is indeed an error 376
margin—and if so, how it is established by the central nervous system. The latter case may be the most promising one since the AAMV was claimed to be a predictor of risk of fall in the elderly (Delignieres et al., 2011a, 2011b), again without further elaboration on this topic. While AAMVx and AAMVy represent the empirical thresholds at which individuals changed their CoP direction in each axis, a more general interpretation applies to AAMVxy as the threshold at which the CoP changes its direction on the XY plane. Interestingly, the threshold AAMVxy also showed a main effect with postural task and somatosensory constraint, but with an unanticipated result: even higher values than those expect from the vector addition offfi Vx and Vy (e.g., FAEO before ischemia: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi j 4:202 C 7:122 j D 8:27cm=s; AAMVxy D 19.43 cm/s). While univariate thresholds (AAMVx, AAMVy) define the semi-axes of an ellipse for the intermittent control of posture based on velocity information, the AAMVxy defines the radius of a circle with much greater area than its elliptical counterpart (Figure 4). Therefore, both AAMVx and AAMVy seem to underestimate the true CoP speed threshold used by the central nervous system and therefore they do not to capture the CoP velocity in the XY plane. Although the center of mass (and consequently CoP) displacement can be separated into orthogonal components, a bivariate threshold is physiologically justified since most body muscles do not act on a single joint axis. This novel finding, not investigated in previous studies using univariate AAMV (Delignieres et al., 2011a, 2011b), suggests the existence of a bivariate threshold to be used by the postural control system that needs further investigation. The statokinematogram variables derived in this work (nHSR, Davg, and Dmax) behave as other CoP variables (e.g. Area and VXYavg), with higher values coming out from more unstable postures (Raymakers et al., 2005). The consistency between this study and previous ones using similar ischemic hypoxia models and CoP signal processing techniques for stabilogram and statokinesigram analysis (Fitzpatrick, Rogers, & McCloskey, 1994; Mauritz & Dietz, 1980; Wang & Lin, 2008) reinforces the external validity of this study. However, none of these referred studies investigated the localization and quantification of HSR, which can be of clinical relevance for investigating falls as following. On one hand, because high CoP velocities should be avoided near the base of support (BoS) boundaries (Riccio, 1993), it is possible to argue whether Davg could actually distinguish those at high risk of falling. On the other hand, because high CoP velocities near the boundaries can be useful to redirect the center of mass toward the egocentric reference of posture, it could be used as a strategy for rehabilitation of people at high risk of falling. Some limitations are discussed, mainly related to the protocol and the proposed method. First, although the period of ischemic hypoxia applied in this study may induce only partial somatosensory loss in healthy subjects (Carpenter et al., 2010), it was sufficient to cause changes in Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
FIGURE 4. Ellipses and circles for center-of-pressure (CoP) velocity, grouped by postural tasks and somatosensory constraint. Top: Comparison of the boundary limits of CoP velocity based on thresholds in the stabilogram (AAMVx, AAMVy) and the statokinesigram (AAMVxy). Notice the effect of postural task and somatosensory constraint in the inner area. Bottom: Comparison of the ellipse and circle areas for the same postural task and somatosensory condition.
somatosensory input by other authors and yielded comparable results to those reported elsewhere (Fitzpatrick et al., 1994; Mauritz & Dietz, 1980; Wang & Lin, 2008). Second, the effect of the signal acquisition duration and sampling frequency on the statokinematogram variables needs to be clarified since both parameters affect others stabilometric parameters (Carpenter, Frank, Winter, & Peysar, 2001; Le Clair & Riach, 1996; van der Kooj, Campbell, & Carpenter, 2011). However, the additional analysis using a high low-pass filtering configuration suggests that the proposed method was robust to increases in cutoff frequency, showing only minor changes in the observed results. Such behavior was not surprising because the frequency content of the CoP signal during undisturbed upright stance is restricted to low frequencies (
Journal of Motor Behavior Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vjmb20
Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss a
a
Fellipe M. Portela & Arthur S. Ferreira a
Laboratory of Computational Simulation and Modeling in Rehabilitation, Centro Universitário Augusto Motta, Rio de Janeiro, Brazil Published online: 19 Jun 2014.
Click for updates To cite this article: Fellipe M. Portela & Arthur S. Ferreira (2014) Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss, Journal of Motor Behavior, 46:5, 369-379, DOI: 10.1080/00222895.2014.916651 To link to this article: http://dx.doi.org/10.1080/00222895.2014.916651
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions
Journal of Motor Behavior, Vol. 46, No. 5, 2014 Copyright © Taylor & Francis Group, LLC
RESEARCH ARTICLE
Kinematic Mapping Reveals Different Spatial Distributions of Center of Pressure High-Speed Regions Under Somatosensory Loss Fellipe M. Portela, Arthur S. Ferreira
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
rio Augusto Motta, Laboratory of Computational Simulation and Modeling in Rehabilitation, Centro Universita Rio de Janeiro, Brazil. ABSTRACT. The spatial distribution of center-of-pressure speed during postural tasks and its changes due to somatosensory constraint (temporary ischemic hypoxia on ankle/feet) were investigated in young, healthy subjects (n D 13). A single high-speed region in the central region of the statokinesigram was observed during postural tasks with full sensory information. A significant increase in the quantity of high-speed regions was observed during ischemia and somatosensory constraint, whereas a significant increase in the quantity of high-speed regions localized more distant to the center of center-of-pressure area occurred under somatosensory constraints, suggesting a redirection of center-ofpressure trajectory to adjust the position of the center of mass with respect to the egocentric reference of balance.
Lin, 2008), and muscle fatigue (Hlavackova & Vuillerme, 2012). Because the combination of CoP position and velocity in a phase-plane plot was regarded as an important tool for assessment of postural control (Bottaro, Yasutake, Nomura, Casadio, & Morasso, 2008; Riley, Benda, GillBody, & Krebs, 1995) and was recently argued as a predictor for risk of fall (Honarvar & Nakashima, 2013) both in an undisturbed upright stance, it is possible that other methods combining kinematic variables provide new insights about the postural control. Despite the hypothesis that a large CoP area may be a natural exploratory behavior for postural control and not a consequence of postural instability (Carpenter, Murnaghan, & Inglis, 2010), higher CoP area and velocity are usually interpreted as worse postural control (Raymakers, Samson, & Verhaar, 2005). Considering the relationships between position and velocity as kinematic variables, an interesting question arises: Where is it necessary to use high CoP velocities for maintaining the upright posture? On the one hand, high CoP velocities should be avoided near the boundaries of the base of support since there may be not enough time for an efficient postural adjustment at these locations (Riccio, 1993). Indeed, the velocity-based postural control was described to occur in the central region of the stabilogram, where CoP velocity reaches its maximal absolute values (Delignieres et al., 2011a), although no evidence was presented regarding this observation. On the other hand, high CoP velocities near the boundaries of the base of support may be necessary to quickly redirect the body’s center of mass toward the egocentric reference of posture in more challenging conditions or after a fall is initiated. Therefore, a method to assess the spatial distribution of regions with high CoP velocity would be valuable for investigation what postural strategies do patients at high risk of falling use. Presently, the spatial distribution of regions with high CoP velocity remains uninvestigated. This work describes the spatial distribution of CoP velocity during the undisturbed upright stance in healthy subjects; a companion method is also introduced for visualization, localization, and quantification of regions with high CoP velocity, namely kinematic mapping. Explicitly stating, this study was motivated by the velocity-
Keywords: center-of-pressure, egocentric constraints, kinematics, postural sway, rehabilitation
D
isplacements of the body’s center of mass while subjects perform postural tasks are described by anteroposterior (AP) and medialateral (ML) coordinates of the body’s center of pressure (CoP; Duarte & Freitas, 2010; Winter, 1995). From the time-sampled CoP coordinates, it is possible to calculate statistical, geometrical, kinematic, structural, and fractal variables related to postural control in one or two spatial dimensions (Baratto, Morasso, Re, & Spada, 2002; Deschamps, Beauchet, Annweiler, Cornu, & Mignardot, 2014; Duarte & Freitas, 2010; Kapteyn et al., 1983). Previous studies (Collins & De Luca, 1993; Duarte & Zatsiorsky, 2002; Juras, Stomba, Fredyk, Sobota, & Bacik, 2008; Zatsiorsky & King, 1998) investigated the postural control in the undisturbed upright stance considering parameters calculated from CoP displacement data. More recent studies emphasized the role of velocity for postural control (Jeka, Kiemel, Creath, Horak, & Peterka, 2004; Masani, Popovic, Nakazawa, Kouzaki, & Nozaki, 2003; Pai & Patton, 1997; Sasagawa, Ushiyamab, Kouzakic, & Kanehisa, 2009) and provided support to the velocity-based theory of an intermittent control of the upright posture (Delignieres, Torre, & Bernard, 2011a). Particularly, kinematic variables (e.g. linear path, displacement, and velocity) are of clinical interest because they reflect postural responses due to ageing (Du Pasquier et al., 2003), chronic ankle instability (Wikstrom, Fournier, & McKeon, 2010), Parkinson’s disease (Mitchell, Collins, De Luca, Burrows, & Lipsitz, 1995), diabetic peripheral neuropathy (Boucher, Teasdale, Courtemanche, Bard, & Fleury, 1995), lower limb somatosensory loss (Wang &
Correspondence address: Arthur de S a Ferreira, Pra¸c a das Na¸c o~es 34, 3 andar, Bonsucesso, Rio de Janeiro, RJ – Brazil, CEP 21041–010. e-mail: [email protected] 369
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
based theory for controlling the undisturbed upright posture, which hypothesizes that CoP velocity is bounded and body sway is left uncheck until a threshold in velocity is reached (Delignieres et al., 2011a). In this theory, the CoP signal and related parameters are used because they reflect the overall process underlying the postural control scheme (Delignieres et al., 2011a; Winter, 2005). The clinical utility of this method regarding identification of spatial patterns under distinct postural challenges was verified with a model of somatosensory loss temporarily induced by an ischemic hypoxia of the lower limbs. It was hypothesized that during unchallenging postural tasks subjects will exhibit one highspeed region (HSR) centered at the CoP area, and that the suppression of somatosensory input would increase both the quantity of HSR as well as the distance away to the center of the CoP area. Method Ethics Statement The Institutional Ethics Committee approved this study before its execution (CAAE-02420912.8.0000.5235). All volunteers were informed about the procedures and gave their written consent to participate in the study. Subjects Thirteen healthy volunteers (eight women; 27 § 5 years old; 69.3 § 8.9 kg; 1.68 § 0.09 m tall) from the academic community agreed to participate in this study. Subjects were admitted after passing both clinical screening and physical examination to exclude musculoskeletal, neurological, and vascular disorders that might compromise their safety and the execution of the protocol. This study was conducted in a silent room with controlled air temperature (21–23 C). Blocks of Postural Tasks Subjects were instructed to step up on the platform barefooted and to keep an undisturbed bipedal position with their arms hanging at their body sides (Figure 1, top panel). They performed a first block lasting 60 s that was preceded by a 20-s period to avoid recording movement artifacts due to stepping up. Combinations of visual and biomechanical constraints characterized the two postural tasks investigated: feet apart (8 cm between heels, 10 angle) with eyes opened (FAEO) and feet together (heels in contact, 10 angle) with eyes closed (FTEC). During tasks with eyes opened, volunteers were instructed to focus on a small target fixed at the eye level on a wall 1.5 m apart from the force platform. Between postural tasks volunteers were allowed to rest during 60 s. These two tasks were performed without wearing the pressure cuffs because they could have augmented the cutaneous sensory information at that level, thus influencing the postural control (Wang & Ling, 2008). 370
FIGURE 1. Positioning for postural tasks and somatosensory constraint. Top: Lower extremity positioning (left: feet apart; right: feet together) before ischemic hypoxia. Bottom: Lower extremity positioning (left: feet apart; right: feet together) during ischemic hypoxia with the sphygmomanometer cuff.
A second block was performed with feet and ankle ischemic hypoxia to generate a temporary somatosensory loss on these body parts (Wang & Lin, 2008). Blood pressure was measure in both arms and legs with an automated device (model G-TECH, BP3AF1-3, Onbo Electronics Co., Shenzen, China) to calculate the ankle-brachial index (group results: right side D 1.2 § 0.1, left side D 1.3 § 0.2) for ruling out subclinical peripheral vascular disease (Al-Qaisi, Nott, King, & Kaddoura, 2009). Two sphygmomanometer cuffs connected by a T tube were positioned above the lateral malleolus of the ankle joint (Figure 1, bottom panel) to interrupt the blood flow and consequently to impair conduction from deep and superficial mechanoreceptors above the level of the cuffs (Horak, Nashner, & Diener, 1990). The cuffs were manually inflated 75 mm Hg above the systolic blood pressure (higher measure from both arms) for 25 min and monitored regularly on both legs. Both postural tasks were repeated after 25 min of ischemic hypoxia with the cuffs still inflated. Following the last task, muscle strength was immediately assessed to evaluate the preservation of efferent innervation. Finally, both cuffs were deflated and the local reactive hyperemia was observed during 5–10 min before discharging the subject of the experiment.
Motor and Sensory Measurements Motor function test was performed based on standard manual testing (Hislop & Montgomery, 1996). Muscle groups assessed included the extensor digitorum longus and brevis, extensor hallucis longus, tibialis anterior, flexor hallucis brevis, lumbricales, and gastrocnemius. Journal of Motor Behavior
Kinematic Mapping of Center of Pressure
affirmative response was sufficient to characterize the respective level of touch-pressure threshold. Instrumentation, Signal Acquisition, and Preprocessing Figure 2 summarizes the computational methods proposed in this study for kinematic mapping of the HSR as implemented in LabVIEW 8.0 (National Instruments, Dallas, TX) running on Windows 7 (Microsoft Corporation, Seattle, WA). Signals were acquired at 100 Hz from analog channels of the force platform (AccuSwayPlus, AMTI, Watertown, MA) and digitalized by a 16-bit converter (§10 V, NI-USB 6210, National Instruments). Digital
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Verbal analog scale was applied every 2 min to assess the level of pain or discomfort from 0 (no discomfort or pain) to 10 (severe discomfort or pain). After 15 min of ischemic hypoxia, cutaneous sensitivity was monitored in a subgroup (n D 5) every 2 min by Semmes-Weinstein monofilament testing (SORRI-Bauru, S~ao Paulo, Brazil) according to the manufacturer’s instructions and previous studies (Wang & Lin, 2008). Briefly, monofilaments were applied in sequence (green: 0.05 g, blue: 0.2 g, violet: 2.0 g, red: 4.0 g, orange: 10.0 g, magenta: 300 g) with the subjects with eyes closed to eliminate visual cues. They were asked to answer “yes” whenever they felt the monofilament. The test was performed three times at the same place such that a single
FIGURE 2. Flow chart for signal acquisition, processing, and analysis of the statokinematogram. Computational methods proposed for generation, preprocessing, and analysis of the statokinematograms. AAMVx D average absolute maximal velocity on X-axis; AAMVy D average absolute maximal velocity on Y-axis; AAMVxy D average absolute maximal velocity on XY plane; CoP D center of pressure; Davg D average Euclidean distance from centroid; Dmax D maximal Euclidean distance from centroid; HSR D high-speed regions; SDVx D standard deviation on X-axis; SDVy D standard deviation on Y-axis; VX D Velocity on X-axis; VXmax D maximal velocity on X-axis; VXY D velocity on XY plane; VXYavg D average velocity on XY plane; VY D velocity on Y-axis; VYmax D maximal velocity on Y-axis.
2014, Vol. 46, No. 5
371
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
signals were stored for off-line processing and used to calculate forces (Fx, Fy, Fz) and moments of force (Mx, My, Mz) with the calibration matrix provided by the platform’s manufacturer as in a previous study (Baracat & de Sa Ferreira, 2013). CoP coordinates were calculated according to the manufacturer’s manual. CoP univariate signals were detrended and had their mean value subtracted before a low-pass digital filter (second order, Butterworth filter, cutoff frequency of 2.5 Hz) was applied in direct and reverse direction (Vieira, Oliveira, & Nadal, 2009). This cutoff frequency is justified simultaneously by the very low amplitude harmonics above 1 Hz in healthy subjects under undisturbed bipedal stance (Vieira et al., 2009) and the need of smooth signals for calculation of time derivatives of CoP displacement (Winter, 2005). CoP signals processed at this stage were used to compute the following variables related to body sway in both ML (X-axis) and AP (Y-axis) directions (Raymakers et al., 2005): standard deviation (SDX, SDY), and maximum velocity (VXmax, VYmax) from stabilograms; elliptical area using the principal component analysis method (Area; Oliveira, Simpson, & Nadal, 1996); and average velocity (Vavg) from statokinesigrams. Calculation of CoP Velocity and Speed CoP velocity was calculated using the finite difference calculus and the central difference method to each coordinate X and Y separately (Vx and Vy, respectively). CoP velocity in the ith sample (equations 1–2) was estimated considering the adjacent samples i–1 and iC1, thus providing a time alignment between the reference position for calculating the CoP velocity and the spatial coordinates of the XY axes (Robertson, Caldwell, Hamill, Kamen, & Whittlesey, 2004; Winter, 2005): xi C 1 ¡ xi ¡ 1 2Dt yi C 1 ¡ yi ¡ 1 Vyi D 2Dt Vxi D
(1) (2)
Where Dt represents the acquisition time interval, i corresponds to the sample index, xi–1 and xiC1 are the CoP coordinates at the X-axis in the samples surrounding the ith sample, and yi–1 and yiC1 are the CoP coordinates at the Yaxis in the samples surrounding the ith sample. The CoP speed in the XY plane (Vxy) for the ith sample was estimated using equation 3: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Vxyi D Vx2i C Vy2i
(3)
Kinematic Mapping of HSR The method of kinematic mapping of HSR is defined as the spatial localization of CoP speed in synchronization with the CoP bivariate time series. The aim of this method 372
is to map areas with CoP speed higher than an empirical threshold (Delignieres et al., 2011b). The thresholds are estimated as the average absolute maximal velocity (AAMV) from a 2-s epoch of Vx and Vy, namely AAMVx and AAMVy respectively. Likewise, a new threshold related to CoP bivariate signal was proposed, namely the average absolute maximal velocity of 2-s epochs of Vxy (AAMVxy). Therefore, regions in which the CoP spatial coordinates have Vxy greater than AAMVxy are denominated HSR. Because AAMVx and AAMVy thresholds are affected by postural tasks (Delignieres et al., 2011b) all thresholds were estimated for each postural task. The statokinematogram is a companion graph for visualization of HSR as obtained with the kinematic mapping. A zero-value bidimensional matrix (bin resolution: 1 £ 1 mm) was generated from the CoP bivariate series. A comparison is then performed between Vxy and AAMVxy: if the ith sample of Vxy estimated for the respective bin coordinate (Xi,Yi) is higher than AAMVxy, its value is summed to the previous value in the respective bin. At the end of this loop the cumulative sum in each bin is divided by the quantity of samples stored on the respective bin to obtain the CoP average speed at the respective bin. Since this map present low spatial resolution and quantization error due to categorization of continuous CoP velocity values into discrete bins, a median spatial filtering with a 3 £ 3 kernel size followed by spatial interpolation with bilinear algorithm (Pedrini & Schwartz, 2008) was performed (Figure 3). See also Movie 1, “Generation of Three-Dimensional Statokinematogram”, provided as an online supplementary file available at www.tandfonline.com/vjmb. Localization of HSR and Quantification of HSR in the Statokinematogram The bidimensional matrix with CoP average speed above AAMVxy was converted to a binary matrix with bins marked either as above (D1) or equal or below (D0) this threshold. In sequence, a two-scan connected-component labeling algorithm (He, Chao, & Suzuki, 2008) was applied to this binary matrix to locate HSRs. Because a variable quantity of HSR was expected, three variables were estimated using the centroid coordinates:
1. nHSR: the quantity of nonconnected HSR (i.e., HSR that do not share bins). This variable represents the quantity of distinct loci where CoP trajectory was changed for stabilization of posture. 2. Davg: the averaged Euclidean distance between all HSR centroids and the center of the CoP area. This variable represents the expected distances from HSR to the CoP area. 3. Dmax: the maximal Euclidean distance among all HSR centroids and the center of the CoP area. This variable represents the extreme distance from HSR to the CoP area, being the maximal value often regarded Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
FIGURE 3. Velocity time series in stabilograms and statokinesigrams, and high-speed regions on the statokinematogram. From top to bottom: Center-of-pressure (CoP) velocity time series from medialateral and anteroposterior stabilograms, and CoP speed time series from statokinesigram (all during feet apart–eyes opened, before ischemic hypoxia). Below is exhibited four statokinematograms from all combinations of postural tasks and somatosensory constraint from the same subject (#5). Dotted lines represent the thresholds for positive and negative axis and higher plane. Colored areas correspond to the high-speed regions (HSR) in the statokinematogram.
as a relevant factor when studying the biomechanics postural control (King & Zatsiorsky, 2002). Statistical Analysis Values in tables are presented as M § SD for continuous variables, median [minimum; maximum] for ordinal variables, and absolute frequency (%) for nominal ones. A twoway repeated-measures analysis of variance was performed to test the null hypothesis of no main or interaction effects related to postural task (levels: FAEO; FTEC) and somatosensory constraint (levels: no ischemia; ischemia) separately for variables of the stabilogram (SDX, SDY, VXmax, VYmax), statokinesigram (Vavg, Area), and statokinematogram (Dmax and Davg). Marginal homogeneity test was 2014, Vol. 46, No. 5
used to test the null hypothesis of equal marginal proportions of nHSR among all four levels. Association between variables was tested using the Spearman rank correlation coefficient (H0: r D 0) under FAEO with no ischemia. Statistical significance was considered at p < .05 with the adjusted p-value for multiple comparisons based on a stepwise rejection procedure (Li, 2008). Statistical analyses were performed in SPSS 17 (SPSS Inc., Chicago, IL). Robustness Analysis Methodological procedures affect the reliability of CoPbased parameters (Ruhe, Feier, & Walker, 2010). Therefore, a separated analysis was conducted to assess whether the kinematic mapping is robust to changes in signal processing 373
F. M. Portela & A. S. Ferreira
techniques. Two important methodological procedures used in this study match the recommendations for optimal reliability (trial duration 60 s and sampling frequency D 100 Hz). Thus is of particular interest the robustness analysis with respect to the low-pass filter cutoff frequency since in this study the selected cutoff frequency was lower than the recommended for processing CoP data (2.5 Hz instead of 10 Hz). The variables of the statokinematogram—nHSR, Davg, and Dmax—were recalculated using a 10 Hz lowpass filter for a subsequent statistical analysis using the same tests described previously for these variables.
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Results Sensory-motor analysis revealed that all subjects presented preserved muscle strength (grade 5) after 25 min of ischemia in every tested muscle. Also, all subjects reported numbness in the lower extremities and pain intensity averaged 4.4 § 1.6. Subjects tested with the monofilaments reported sensitivity at the violet-colored monofilament, interpreted as increased touch-pressure threshold. An overall significant difference in marginal proportions (p D .001) was observed among the four levels in nHSR (Table 1). Paired analysis of nHSR revealed significant marginal differences for somatosensory constraint during FTEC (p D .011), as well as for postural tasks during ischemia (p D .018). A significant interaction between postural task and somatosensory constraint was observed for Davg, Wilks’ L D .708, F(1, 12) D 4.961, p D .046, h2 D .292.
Dmax was significantly higher during FTEC, Wilks’ L D .485, F(1, 12) D 12.718, p D .004, h2 D .515; and somatosensory constraint, Wilks’ L D .586, F(1, 12) D 8.470, p D .013, h2 D .414, but without significant interaction (p D .053). AAMVx was significantly higher during FTEC, Wilks’ L D .211, F(1, 12) D 44.890, p < .001, h2 D .789, but not significantly different under somatosensory constraint, Wilks’ L D .831, F(1, 12) D 2.448, p D .144, h2 D .169. AAMVy was significantly higher during FTEC, Wilks’ L D .219, F(1, 12) D 42.778, p < .001, h2 D .781, and somatosensory constraint, Wilks’ L D .157, F(1, 12) D 64.521, p < .001, h2 D .843, but without significant interaction (p D .114). Finally, AAMVxy was significantly higher during FTEC, Wilks’ L D .207, F(1, 12) D 45.921, p < .001, h2 D .793, and somatosensory constraint, Wilks’ L D .195, F(1, 12) D 49.692, p < .001, h2 D .805, but without significant interaction (p D .141). Area was significantly higher during FTEC, Wilks’ L D .300, F(1, 12) D 27.967, p < .001, h2 D .700, and somatosensory constraint, Wilks’ L D .541, F(1, 12) D 10.166, p D .008, h2 D .459, but without significant interaction (p D .068). Vavg was significantly higher during FTEC, Wilks’ L D .173, F(1, 12) D 57.412, p < .001, h2 D .827, and somatosensory constraint, Wilks’ L D .299, F(1, 12) D 28.170, p < .001, h2 D .701, but without significant interaction (p D .190). SDx was significantly higher during FTEC, Wilks’ L D .170, F(1, 12) D 58.771, p < .001, h2 D .830, but not somatosensory constraint, Wilks’ L D .997, F(1, 12) D
TABLE 1. Comparative Analysis of Stabilometric Parameters According to the Postural Task and Somatosensory Constraints on Variables From the Stabilogram, Statokinesigram, and Statokinematogram Feet apart, eyes opened Variable Velocity thresholds AAMVx (cm/s) AAMVy (cm/s) AAMVxy (cm/s) Statokinematogram nHSR Davg (mm) Dmax (mm) Statokinesigram Area (cm2) Vavg (cm/s) Stabilogram SDX (cm) SDY (cm) VXmax (cm/s) VYmax (cm/s)
Feet together, eyes closed
Before hypoxia
During hypoxia
Before hypoxia
During hypoxia
p Values
4.20 § 0.79 7.12 § 1.55 19.43 § 3.44
4.22 § 0.72 8.49 § 2.07 23.22 § 5.64
8.98 § 2.98 12.42 § 4.01 35.62 § 10.46
9.68 § 3.48 15.31 § 4.78 43.17 § 13.37
< .001a; .144b < .001a; < .001b < .001a; < .001b
1 [1, 2] 1.67 § 1.56 2.21 § 2.41
1 [1, 4] 2.38 § 2.17 3.53 § 3.50
1 [1, 4] 3.04 § 2.26 4.69 § 4.12
3 [1, 5] 6.03 § 2.55 9.38 § 4.22
.001d .046d .004a; .013b
0.37 § 0.19 0.42 § 0.08
0.42 § 0.26 0.47 § 0.09
1.30 § 0.77 0.80 § 0.24
1.78 § 1.05 0.93 § 0.27
< .001a; .008b < .001a; < .001b
0.12 § 0.03 0.26 § 0.07 1.23 § 0.48 2.32 § 1.08
0.11 § 0.04 0.31 § 0.11 1.03 § 0.44 2.44 § 1.14
0.28 § 0.09 0.38 § 0.14 2.48 § 1.23 3.50 § 1.89
0.29 § 0.10 0.51 § 0.17 3.15 § 1.74 4.11 § 2.20
< .001a; .857b .001a; .001b .033c .018a; .408b
Note. Data format is M § SD or median [minimum, maximum]. a Main effect for postural task. bMain effect for somatosensory constraint. cInteraction effect between postural task and somatosensory constraint. d Friedman’s test.
374
Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
0.034, p D .857, h2 D .003. SDy was significantly higher during FTEC, Wilks’ L D .365, F(1, 12) D 20.870, p D .001, h2 D .635, and somatosensory constraint, Wilks’ L D .415, F(1, 12) D 16.887, p D .001, h2 D .585, but without significant interaction (p D .066). A significant interaction between postural task and somatosensory constraint was observed for VXmax, Wilks’ L D .674, F(1, 12) D 5.812, p D .033, h2 D .326. VYmax was significantly higher during FTEC, Wilks’ L D .614, F(1, 12) D 7.544, p D .018, h2 D .386, but not somatosensory constraint, Wilks’ L D .942, F(1, 12) D 0.736, p D .408, h2 D .058. Area was positively correlated to AAMVx (r D .569; p D .021), AAMVy (r D .664; p D .007), and AAMVxy (r D .641; p D .009). Variable nHSR was positively correlated to Davg (r D .804; p < .001) and Dmax (r D .804; p < .001). Conversely, Area was not significantly correlated to nHSR (r D .379; p D .101), Davg (r D .426; p D .073), or Dmax (r D .426; p D .073). Table 2 exhibits the results of the robustness analysis. Parameters nHSR, Davg, and Dmax were higher and generally more disperse if calculated with the 10 Hz cutoff frequency. Using this higher cutoff frequency yielded a similar behavior with a minor deterioration of the observed main and interaction effects. An overall significant difference in marginal proportions (p D .017) was maintained among the four levels in nHSR. Paired analysis of nHSR revealed again significant marginal differences for postural tasks during ischemia (p D .026) but not for somatosensory constraint during FTEC (p D .912). Davg was significantly higher during FTEC, Wilks’ L D .542, F(1, 12) D 10.144, p D .008, h2 D .458, but not under somatosensory constraint, Wilks’ L D .850, F(1, 12) D 2.121, p D .171, h2 D .150, without significant interaction (p D .774). Likewise, Dmax was significantly higher during FTEC, Wilks’ L D .500, F(1, 12) D 11.982, p D .005, h2 D .500, but not under
somatosensory constraint, Wilks’ L D .794, F(1, 12) D 3.109, p D .103, h2 D .206, again without significant interaction (p D .214). Discussion This work described the spatial distribution of CoP speed during the undisturbed upright stance in healthy subjects using the proposed method of kinematic mapping of HSR. The clinical utility of this method was verified with a model of somatosensory loss temporarily induced by ischemic hypoxia of the lower limbs, a method well-known for generating postural challenges in case of combined visual, biomechanical, and somatosensory constraints (Horak et al., 1990; Wang & Ling, 2008). The main results of this study comprise (a) the undisturbed upright stance is characterized by one HSR nearly centered at the CoP area, and (b) more challenging postural tasks increased both the quantity of HSR as well as their distance from the center of the CoP area. This is the first work to objectively localize and quantify HSRs, and most importantly, to show how their spatial distribution changes under somatosensory constraints. In other words, the observed results suggest that different velocity-based strategies for redirecting the body’s center of mass to the reference position area were used under more challenging postural tasks. Collectively, these results support that the kinematic mapping was sensitive to changes in postural task and somatosensory constraints. Most importantly, these results emphasize that the combination of CoP position and velocity provided by the kinematic mapping allows the assessment of the postural strategies used under different somatosensory and visual/ biomechanical constraints. Our hypothesis that a single HSR appears in the statokinematogram during postural tasks with full sensory
TABLE 2. Robustness Analysis of the Parameters From the Statokinematogram at Different Low-Pass Filter Cut-Off Frequencies Feet apart, eyes opened Variable nHSR 2.5 Hz 10.0 Hz Davg (mm) 2.5 Hz 10.0 Hz Dmax (mm) 2.5 Hz 10.0 Hz
Feet together, eyes closed
Before hypoxia
During hypoxia
Before hypoxia
During hypoxia
p Values
1 [1, 2] 2 [1, 4]
1 [1, 4] 1 [1, 4]
1 [1, 4] 3 [1, 7]
3 [1, 5] 3 [1, 9]
1.67 § 1.56 2.91 § 1.67
2.38 § 2.17 3.60 § 3.50
3.04 § 2.26 5.02 § 2.10
6.03 § 2.55 6.12 § 3.07
.046c .008a, .171b
2.21 § 2.41 4.41 § 2.90
3.53 § 3.50 4.62 § 4.04
4.69 § 4.12 7.19 § 3.56
9.38 § 4.22 9.73 § 4.92
.004a, .013b .005a, .103b
.001d .017d
Note. Data format is M § SD or median [minimum, maximum]. a Main effect for postural task. bMain effect for somatosensory constraint. cInteraction effect between postural task and somatosensory constraint. d Friedman’s test.
2014, Vol. 46, No. 5
375
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
F. M. Portela & A. S. Ferreira
information was confirmed. Although the velocity-based postural control has been suggested to occur in the central region of the stabilogram (CoP univariate analysis; Delignieres et al., 2011a), this is the first study to reveal the spatial localization of HSR in terms of CoP bivariate analysis. Also, our hypothesis of increased nHSR due to more challenging conditions was also confirmed for both postural task and somatosensory constraint during FTEC. Although the reason for occurrence of more HSR could not be determined in this study, we argued that such observation reflects the need for more changes in CoP trajectory redirecting the center of mass to maintain an upright posture under limited sensory input. Likewise, our hypothesis that suppression of sensory input increases the distance from HSR to the center of CoP area was confirmed. The interaction effect observed for Davg indicated that the effect of the absence of vision and change biomechanics was greater during ischemia for this variable. The fact that an increased quantity of HSR was localized more distant to the center of CoP area suggests that redirecting of CoP trajectory was more necessary to adjust the position of the center of mass with respect to the egocentric reference of balance. Although it may be argued that increased statokinematogram variables are expected due to also increased CoP area between postural tasks, the lack of significant correlations between Area and all statokinematogram variables do not favors this argument. Therefore, the observed changes is spatial distribution seems to be not explained by changes in CoP area and this fact needs to be further investigated. AAMVx showed a significant main effect for postural task, but not for somatosensory constraint, while AAMVy showed a significance main effect for both factors. This suggest that the AP threshold for redirecting the CoP was adjusted to fit the new condition of somatosensory constraint, which is in agreement with the sensory reweighting during limited sensory information (Redfern, Yardley, & Bronstein, 2001). Despite small differences between protocols and signal processing techniques, the calculated AAMVx and AAMVy in this study show similar behavior to those previously observed (Delignieres et al., 2011b) in relation to the group average values and the main effect under no visual input. However, no explanation was provided by the authors (Delignieres et al., 2011a, 2011b) to the observed increase in AAMVx and AAMVy as estimated with and without visual input. Considering the significant, positive correlation coefficients observed between Area and all three thresholds, our explanation is that a considerable amount of increase in CoP velocity threshold is explained by an increase in CoP area. The remaining increase in threshold may be interpreted as an error margin to tolerate higher CoP displacements under conditions of limited sensory input. Further studies should investigate if there is a causal relationship between CoP area and those empirical velocity thresholds, and if the increase in AAMV that is not explained by increase in Area is indeed an error 376
margin—and if so, how it is established by the central nervous system. The latter case may be the most promising one since the AAMV was claimed to be a predictor of risk of fall in the elderly (Delignieres et al., 2011a, 2011b), again without further elaboration on this topic. While AAMVx and AAMVy represent the empirical thresholds at which individuals changed their CoP direction in each axis, a more general interpretation applies to AAMVxy as the threshold at which the CoP changes its direction on the XY plane. Interestingly, the threshold AAMVxy also showed a main effect with postural task and somatosensory constraint, but with an unanticipated result: even higher values than those expect from the vector addition offfi Vx and Vy (e.g., FAEO before ischemia: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi j 4:202 C 7:122 j D 8:27cm=s; AAMVxy D 19.43 cm/s). While univariate thresholds (AAMVx, AAMVy) define the semi-axes of an ellipse for the intermittent control of posture based on velocity information, the AAMVxy defines the radius of a circle with much greater area than its elliptical counterpart (Figure 4). Therefore, both AAMVx and AAMVy seem to underestimate the true CoP speed threshold used by the central nervous system and therefore they do not to capture the CoP velocity in the XY plane. Although the center of mass (and consequently CoP) displacement can be separated into orthogonal components, a bivariate threshold is physiologically justified since most body muscles do not act on a single joint axis. This novel finding, not investigated in previous studies using univariate AAMV (Delignieres et al., 2011a, 2011b), suggests the existence of a bivariate threshold to be used by the postural control system that needs further investigation. The statokinematogram variables derived in this work (nHSR, Davg, and Dmax) behave as other CoP variables (e.g. Area and VXYavg), with higher values coming out from more unstable postures (Raymakers et al., 2005). The consistency between this study and previous ones using similar ischemic hypoxia models and CoP signal processing techniques for stabilogram and statokinesigram analysis (Fitzpatrick, Rogers, & McCloskey, 1994; Mauritz & Dietz, 1980; Wang & Lin, 2008) reinforces the external validity of this study. However, none of these referred studies investigated the localization and quantification of HSR, which can be of clinical relevance for investigating falls as following. On one hand, because high CoP velocities should be avoided near the base of support (BoS) boundaries (Riccio, 1993), it is possible to argue whether Davg could actually distinguish those at high risk of falling. On the other hand, because high CoP velocities near the boundaries can be useful to redirect the center of mass toward the egocentric reference of posture, it could be used as a strategy for rehabilitation of people at high risk of falling. Some limitations are discussed, mainly related to the protocol and the proposed method. First, although the period of ischemic hypoxia applied in this study may induce only partial somatosensory loss in healthy subjects (Carpenter et al., 2010), it was sufficient to cause changes in Journal of Motor Behavior
Downloaded by [Eindhoven Technical University] at 23:25 13 February 2015
Kinematic Mapping of Center of Pressure
FIGURE 4. Ellipses and circles for center-of-pressure (CoP) velocity, grouped by postural tasks and somatosensory constraint. Top: Comparison of the boundary limits of CoP velocity based on thresholds in the stabilogram (AAMVx, AAMVy) and the statokinesigram (AAMVxy). Notice the effect of postural task and somatosensory constraint in the inner area. Bottom: Comparison of the ellipse and circle areas for the same postural task and somatosensory condition.
somatosensory input by other authors and yielded comparable results to those reported elsewhere (Fitzpatrick et al., 1994; Mauritz & Dietz, 1980; Wang & Lin, 2008). Second, the effect of the signal acquisition duration and sampling frequency on the statokinematogram variables needs to be clarified since both parameters affect others stabilometric parameters (Carpenter, Frank, Winter, & Peysar, 2001; Le Clair & Riach, 1996; van der Kooj, Campbell, & Carpenter, 2011). However, the additional analysis using a high low-pass filtering configuration suggests that the proposed method was robust to increases in cutoff frequency, showing only minor changes in the observed results. Such behavior was not surprising because the frequency content of the CoP signal during undisturbed upright stance is restricted to low frequencies (
