A

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Description

Athletic skill level is reflected in body sway: A test case for accelometry in combination with stochastic dynamics

Tags

Transcript

Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics
Claudine J.C. Lamoth
a,d,
*, Rob C. van Lummel
b,c
, Peter J. Beek
a
a
Research Institute MOVE, Faculty of Human Movement Sciences, VU University Amsterdam, The Netherlands
b
McRoberts BV, The Hague, The Netherlands
c
Haagse Hogeschool Human Kinetic Technology
,
The Hague, The Netherlands
d
Center for Human Movement Sciences, University Medical Center Groningen, University of Groningen, The Netherlands
1. Introduction
Adequate postural control is essential for normal daily activityand requires the integration of visual, proprioceptive andvestibular information. The degree to which individuals rely onthose information sources depends on task difﬁculty, cognitiveload [32], motor skill [2,28] and pathology [19,24,29]. In the clinic,
posturography is becoming increasingly important, both fordiagnostic and monitoring purposes and for the evaluation of interventions[6,9].Changesanddifferencesinposturalcontrolare
typically examined using average scalar parameters such as meansway path,sway area, sway velocity,or mean frequency.However,precise deﬁnitions and optimal selections of discriminative swayparameters are under continuous dispute [1,20].
Alternatively, measures derived from the theory of stochasticdynamics may be usefully employed to quantify the time-varyingstructure of postural sway patterns. Even without any externalperturbation during quiet upright stance, an apparently randomandirregularsmallamplitudebodyswayiscontinuouslypresent.The insight that this irregular sway pattern is not simply theproduct of noise, but rather results from a complex interplay of non-linear deterministic and random components, has become aprominent theme in fundamental studies of postural control[4,5,16,23]. In this approach, sway ﬂuctuations are exploited as a
window into underlying control structures. A wide variety of methods has been used for this purpose, including recurrenceanalysis[23],detrendedﬂuctuationanalysis[17],sampleentropy
[21], diffusion coefﬁcients [5], and Lyapunov exponents [26].
Although conceptually different, all these measures are based onthe assumption that the effectiveness of postural control isin some way reﬂected in the dynamic characteristics of swaypatterns, and that differences in those characteristics reﬂectdifferences in postural control. This development has led to theapplication of the aforementioned methods in applied ﬁeldssuch as the clinic and sports medicine, both for diagnostic andprognostic purposes.Postural variability can be scored along a continuum with‘normal’ or ‘healthy’ variability positioned somewhere between
Gait & Posture xxx (2009) xxx–xxx
A R T I C L E I N F O
Article history:
Received 1 February 2008
Received in revised form 25 September 2008
Accepted 2 December 2008
Keywords:
Postural swayVariabilityAccelerometryNon-linear analysisAthletic skill
A B S T R A C T
Recent studies on postural control have shown that the variability of body sway during quiet standingmayprovidevaluableinformationtocharacterizechangesinposturalcontrolduetoage,pathology,skillandtask.Theaimofthepresentstudywastodetermine–asspadeworkforpossibleclinicalapplications– whether body sway measured with a three-axial accelerometer at the trunk can differentiate betweenthree healthy young populations that differ in athletic skill level. The three groups in question (groupsize:
n
= 22) consisted of regular bachelor students, physical education students and physical educationstudents specialized in gymnastics. Data were recorded during tandem stance with eyes open or closedand while standing on foam. The acceleration time-series were analysed in anteriorposterior andmediolateral direction. Differences in postural control were quantiﬁed in terms of variability, spectralproperties and stochastic dynamical measures, i.c., regularity (sample entropy, long-range correlations)and local stability (largest Lyapunov exponent). The results were clear-cut. Standing with eyes closedand on foam increased variability. Compared to standing with eyes open, standing with eyes closedresulted in less regular sway patterns but with greater local stability, whereas standing on foam had anopposite effect. With greater gymnastic skills, acceleration time-series were less variable, less regularand more stable. These results imply that quantifying the stochastic-dynamical structure of posturalsway using ambulant accelerometry may provide a useful diagnostic tool.
2008 Elsevier B.V. All rights reserved.
* Corresponding author at: Center for Human Movement Sciences, UniversityMedical Center Groningen, University of Groningen, Ant. Deusinglaan 1, 9713 AVGroninging, The Netherlands. Tel.: +31 50 3639036.
E-mail address:
C.J.C.Lamoth@med.umcg.nl (Claudine J.C. Lamoth).
G Model
GAIPOS-2711; No of Pages 6
Please cite this article in press as: Lamoth CJC, et al. Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics. Gait Posture (2009), doi:10.1016/j.gaitpost.2008.12.006
Contents lists available at ScienceDirect
Gait & Posture
journal homepage: www.elsevier.com/locate/gaitpost
0966-6362/$ – see front matter
2008 Elsevier B.V. All rights reserved.doi:10.1016/j.gaitpost.2008.12.006
both extremes. This view is in line with the notion that health ischaracterized by ‘organized’ variability, while disease is deﬁnedby loss of complexity, increased regularity, and changes in thestructureofvariability[10].Increasedregularityof posturalswayhas been reported for several patient groups, including strokepatients [24], athletes with sports related concussions [3],
patientswithParkinson’sdisease[29],andchildrenwithCerebralPalsy [7]. Pathological sway patterns – characterized by higherregularityandreducedlocalstabilityandcomplexity–arelocatedat one end of the continuum. In general, such characteristics aredeemed to reﬂect a less efﬁcient and less automatized form of postural control that is less adaptable and more susceptible toexternal perturbations. In contrast, sway patterns obtained fromindividuals with superior balance skills such as gymnasts anddancersarelocatedattheotherendofthecontinuumandarelessregular, locally more stable, and more complex. In general, suchcharacteristicsareconsideredtoindicateamoreautomatizedandmoreadaptiveformofposturalcontrolthatisresilienttoexternalperturbations [14,28].
From this conceptual and methodological background, quanti-ﬁcation of the stochastic-dynamical structure of postural swaypatterns may well enhance fundamental and clincial posturogra-phy. The most common assessment of balance control is based ontheanalysisofcentreofpressure(COP)time-seriesregisteredwitha (single or dual) force plate. In the present study, however, weused accelerometry. The use of body ﬁxed accelerometers tomeasure postural sway is relatively new. Sensors can be attachedto the body close to the centre of body mass in order to measurebody motion. For clinical purposes, accelerometric systems areparticularly useful, as they are small and easy to use. Furthermore,they can be readily used in both controlled (rehabilitationlaboratories) anddailylifecircumstanceswithminimalawarenessof the measuring process on the part of the subject. Preliminaryevidence suggests that the root mean square amplitude of anaccelerometer signal can quantify balance during stance, anddiscriminate between normal elderly adults and idiopathic fallers[12,13,15].
The aim of the present study was to examine if stochastic-dynamical analyses of body sway acceleration signals candiscriminate the postural sway patterns of three populations thatdiffer with respect to their athletic skill level. In addition, weexamined if task difﬁculty had a differential effect in the threegroups, by obtaining sway measures during tandem stance witheyes open (EO), eyes closed (EC), and while standing on foam.Based on previous research examining COP trajectories, weexpected visual deprivation to increase regularity and to decreaselocal stability [8,24], but less so in the physical education groups
than in the regular student group, because gymnasts are known torely less on visual information than non-gymnasts in unstablestance [31]. Finally, we expected the skilled gymnasts to beaffectedlesswhenstandingonfoam,sincegymnastsareparticulartrained to use sensory/proprioceptive information [32]. We usedthe same set of dynamic outcome measures as in the clinicalstudy of Roerdink et al. [24], because the study was intended todetermine whether the combination of accelerometry andstochastic-dynamical off-line analyses could provide a usefuldiagnostic tool.
2. Methods
2.1. Participants
Three groups of 22 subjects, differing in athletic skill level,participated in the experiment: regular undergraduate studentswho did not engage for more than 3 h a week in sport activities,physical education undergraduates pursuing a teacher degree notspeciﬁcallytrainedingymnastics,andphysicaleducationstudentswho specialized in gymnastics. All subjects were healthy and hadno orthopedic or neurological problems that could affect posturalcontrol. All subjects provided written informed consent toparticipate in the study.
2.2. Instrumentation
Accelerations during standing were measured with a tri-axialpiezo-capacitive accelerometer (DynaPort
1
MiniMod, McRobertsBV, The Hague, the Netherlands). The acceleration module(64 mm
64 mm
13 mm) was ﬁxed to the body with an elasticbeltnearthecentreofmassattheleveloflumbarsegmentL3.Datawere sampled at 100 Hz and stored on an SD-card. A radiographicremote control unit was used to start and stop the measurement,and to mark off each measurement episode.
2.3. Procedure and data analysis
Participants stood barefoot in a two legged tandem stance on a10 cm wide and 18 mm thick plywood strip. The experimentalconditions were standing with eyes open (EO), standing with eyesclosed (EC), and standing on foam (FO). Data collection wasinitiated when participants indicated they stood stable and wereready to begin. Each trial lasted 30 s.Anteriorposterior (AP) and mediolateral (ML) accelerationtime-series were analysed. Due to the horizontal tilt of theaccelerometers (e.g., caused by the curvature of the low back) thegravity component may induce variability unrelated to balancecontrol. All data were corrected for this horizontal tilt [15]. Inaddition, a high pass third order Butterworth bidirectional ﬁlterwasappliedwithacut-offfrequencyof0.016 Hztocorrectforslowdriftsduringstandingandalow-passﬁlterwithacut-offfrequency20 Hz to eliminate low amplitude measurement noise [15,24].
1
Sway variability was quantiﬁed by calculating the root meansquares (RMS) of the AP and ML acceleration time-series. Thefrequency of the time-series was determined by calculating themeanpowerfrequency(MPF)fromthepowerfrequencyspectrum,using a 3000 samples Hamming window without overlap, andlinear trends were removed. Postural sway dynamics wasquantiﬁed by means of the sample entropy (SEn) [21], the scalingexponent
a
[17], and the largest Lyapunov exponent (
l
max
) [26],which are brieﬂy described below and more extensively in theAppendix.SE
n
indexes the regularity or predictability of a time-series andis used to analyse complexstochastic systems, which by deﬁnitionis composed of both deterministic and random components [18].SEn is deﬁned as the negative natural logarithm of an estimate of the conditional probability that epochs of length
m
(in this study
m
= 3) match pointwise within a tolerance
r
also match at the nextpoint. Small SEn values are associated with great regularity whilelarge SEn values represent a small chance of similar data beingrepeated, that is, great irregularity. Before calculating SEn the datawere ﬁrst normalized to unit variance, rendering the outcomescale-independent. To quantify the extent to which a recordedtime-seriesexhibitlong-rangecorrelations(i.e.,similarpatternsof variation across multiple time scales) detrended ﬂuctuationsanalysis(DFA)wasapplied.Iftheoutcomeparameter
a
isbetween0.5 and 1, this indicates the presence of long-range power-lawcorrelations in the time-series; that is for white noise, alpha is0.5, for 1/
f
noise it is 1, and for Brownian noise it is 1.5. Finally, thesystem’s resistance to small internal perturbations, such as the
1
Note that ﬁltering may have subtle effects on the non-linear structure detectedinthedata;however,thepotentialeffectoflow-passﬁlteringwaslimitedgiventhat95% of the power of the time-series was located at the lower frequencies.
C.J.C. Lamoth et al./Gait & Posture xxx (2009) xxx–xxx
2
G Model
GAIPOS-2711; No of Pages 6
Please cite this article in press as: Lamoth CJC, et al. Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics. Gait Posture (2009), doi:10.1016/j.gaitpost.2008.12.006
natural sway ﬂuctuations present while standing upright, wasassessed by means of
l
max
[26]. If
l
max
is negative, then anyperturbation exponentially damps out and initially close trajec-tories remain close. In contrast, for positive
l
max
, nearby pointsdiverge as time evolves and produce instability, that is, thedistance between trajectories increases exponentially.To test for spurious effects surrogate data were created. Datawere time-randomized, by randomly selecting samples, thusdestroying the data’s temporal correlation while preserving thedata’s statistics (see Fig. 1 for an example). The absence of temporal correlations is reﬂected in a
a
close to 0.5 and largevalues for SEn. For time-randomized data no embeddingdimension can be estimated, rendering caluculation of
l
max
impossible. Phase-randomized surrogate data were obtained byrandomizing the data’s Fourier phases. This procedure does notalter the spectral power distribution and thus preserves thedata’s auto-correlation function. Hence, scaling exponents of phase-randomized and srcinal data are equal, whereas esti-mates of SEn and
l
max
are increased in the surrogate datacompared to the srcinal data. All data were analysed usingMatlab version 7.
2.4. Statistics
Repeated measures ANOVAs were conducted on all dependentmeasures with GROUP (regular, physical education and gymnasticstudents) as between factor and CONDITION (standing with eyesopen, eyes closed and on foam) as within factor.For signiﬁcant main effects post-hoc
t
-tests were applied todetermine signiﬁcant differences. To evaluate the strength of signi-ﬁcant effects, effect sizes were calculated using
r
¼
ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
t
2
=
t
2
þ
d f
q
,with a 0.24
>
r
>
0.35 indicating a medium effect and
r
>
0.35 alarge effect [27]. Differences between surrogate and srcinal datawere tested using a paired
t
-test, with all conditions and groupscollapsed.
3. Results
Age, weight, length, and body mass index did not differsigniﬁcantly between the three groups (Table 1). The studentgroup participated signiﬁcantly (
p
<
0.001) less in sports activitiesthan the physical education and gymnastic students. Althoughmorefemalesparticipatedinthegymnasticgroupandinthesportsactivities group than in the regular student group, we found nosigniﬁcant gender differences within any of the groups (see also[11,25]).
3.1. Condition
Table 2 provides an overview of all signiﬁcant main effects of condition. Post hoc analysis revealed a signiﬁcantly smaller RMSfor both directions when standing with EO (
t
= 6.50,
p =
0.001,
r =
0.63) than when standing with EC and on foam (
t
= 8.32,
p =
0.001,
r =
0.72). RMS values of ML and AP acceleration of tandemstancewithECdidnotdiffersigniﬁcantlyfromstandingonfoam (Fig. 2). MPF was higher when standing with EC than whenstanding with EO
(t
= 2.22,
p =
0.03,
r =
0.19) and on foam (
t
= 5.28,
p
<
0.001,
r =
0.42). MPF was lower for standing on foam than forstanding with EO (
t
= 3.08,
p =
0.003,
r =
0.26).The main effect of condition on the regularity of the bodyacceleration revealed smaller SEn values of the AP time-serieswhen standing on foam compared to standing with EO and EC(
t
= 3.66,
p
<
0.001,
r
= 0.41).The scaling exponent
a
of AP acceleration time-series wassigniﬁcantly smaller when standing with EC (
t
= 1.92,
p =
0.04,
r =
0.25) than when standing with EO, but was larger whenstandingonfoam(
t
= 5.71,
p
<
0.001,
r =
0.58)thanwhenstandingwith EO. For ML accelerations, however,
a
was signiﬁcantlysmaller when standing on foam than when standing with EO(
t
= 2.66,
p =
0.01,
r =
0.31)orstandingwithEC(
t
= 3.78,
p
<
0.001,
r =
0.42).The
a
forstandingwithEOwasnotsigniﬁcantlydifferentfrom that of EC.
Fig. 1.
Example srcinal (ﬁltered; see text) time-series of mediolateral accelerations of a regular student, a physical education and a gymnastic student when standing witheyes open (left panels), as well as the time randomized (middle panels) and phase randomized counterparts of the time-series in question (right panels).
Table 1
Descriptive characteristics of the participants, means (standard deviations).Subjects Male Female Age (years) Weight (Kg) Length (cm) BMI Sport (hr/wk)Regular students 11 11 21.9 (2.6) 69.5 (9.4) 177.3 (7.6) 22.9 (2.4) 1.2 (1.2)Physical education 8 14 19.3 (2.6) 66.8 (8.5) 174.6 (8.5) 21.7 (1.8) 6.4 (2.8)Gymnastics 4 18 21.3 (3.9) 65.4 (9.0) 170.7 (7.5) 22.3 (3.0) 5.6 (5.1)
C.J.C. Lamoth et al./Gait & Posture xxx (2009) xxx–xxx
3
G Model
GAIPOS-2711; No of Pages 6
Please cite this article in press as: Lamoth CJC, et al. Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics. Gait Posture (2009), doi:10.1016/j.gaitpost.2008.12.006
Standing on foam increased the local stability in AP directioncompared to standing with EO (
t
= 4.48,
p
<
0.001,
r =
0.40) and EC(
t
= 4.80,
p
<
0.001,
r =
0.51), as indicated by a signiﬁcant increasein
l
max
.
3.2. Group differences
A main effect of group was observed for RMS and SEn valuesof both AP and ML accelerations, as well as for MPF and
l
max
of APaccelerations (Table 2). In contrast to the main effect of condition,the scalingexponent
a
did not differsigniﬁcantly betweengroups.No signiﬁcant Group
Condition interactions were observed.Post hoc analysis revealed that for the AP direction RMS wassmaller for the gymnastic group than for the physical educationgroup (
t
= 2.36,
p =
0.02,
r
= 0.21) and the regular studentgroup (
t
= 4.12,
p
<
0.001,
r
= 0.38), whereas RMS of the physicaleducation group was smaller than that of the regular students(
t
= 2.38,
p
= 0.02,
r
= 0.23). The gymnastic group also differed inRMS value in the ML direction compared to the physical educationgroup (
t
= 3.31,
p =
0.001,
r
= 0.28) and the regular students(
t
= 4.32,
p
<
0.001,
r
= 0.35).MPF of AP accelerations was signiﬁcantly larger in thegymnastic group compared to the regular students (
t
= 3.01,
p =
0.003,
r =
0.26) and the physical education group (
t
= 3.61,
p =
0.001,
r =
0.30).Localstabilityinbothdirections (APand ML),i.e.,as indexedbythe inverse of
l
max
, was signiﬁcantly larger for the gymnasticgroup compared to the regular students for AP (
t
= 2.06,
p
= 0.04,
r
= 0.18) and ML directions (
t
= 2.72,
p =
0.007,
r
= 0.23), whereasno signiﬁcant differences were observed between the gymnasticand physical education students, nor between physical educationand regular students.All three groups differed signiﬁcantly form each other withrespect to the regularity of AP and ML accelerations in thatregularity decreased with athletic skill. SEn of AP acceleration wassmaller for regular students than for physical education students(
t
= 2.02,
p
= 0.014,
r
= 0.24), and gymnastic students (
t
= 4.20,
p
<
0.001,
r
= 0.35). In addition, SEn was smaller for physicaleducationstudentsthanforgymnasticstudents(
t
= 2.51,
p =
0.014,
r
= 0.22). Similarly, SEn of ML was signiﬁcantly smaller in theregular education group than in the physical education group(
t
= 1.87,
p
= 0.05,
r
= 0.19) and the gymnastic group (
t
= 4.99,
p
<
0.001,
r
= 0.40), and signiﬁcantlylarger in thegymnastic groupthan in the physical education group (
t
= 3.24,
p
<
0.001,
r
= 0.27).
3.3. Surrogate analysis
Long-range correlations disappeared for the time randomizeddata (
t
= 43.05,
p
<
0.001) but not for the phase randomized data(Fig. 3). Sample entropy was signiﬁcantly larger for the timerandomized data (
t
= 129.84,
p
<
0.001) and to a lesser extent forthe phase randomized (
t
= 13.14,
p
<
0.01). Local stability wassigniﬁcantly greater for the phase randomized data compared tothe srcinal data (
t
= 62.9,
p
<
0.001).
4. Discussion
The goal of the present study was to determine whether bodysway as measured with a three-axial accelerometer combinedwith off-line stochastic-dynamical time-series analyses could
Table 2
Main effects of condition (standing with eyes open, eyes closed and on a foamsurface) and of group (regular students, physical educations students, gymnasticstudents) on mediolateral (ML) and anteriorposterior (AP) variability (RMS), meanpower frequency (MPF), long-range correlations (
a
), regularity (SEn) and localstability (
l
max
). No signﬁcant interactions of condition with group were observed.Direction Condition Group
F
(2,126)
p F
(2,63)
p
RMS AP 34.25
<
0.001 7.60 0.001ML 34.32
<
0.001 7.62 0.001MPF AP 13.61
<
0.001 0.68
ns
ML 4.29
ns
5.22 0.008
a
AP 16.32
<
0.001 0.06
ns
ML 9.89
<
0.001 0.70
ns
l
max
AP 9.05
<
0.001 1.21
ns
ML 0.54
ns
3.17 0.049SEn AP 6.67 0.002 5.54 0.006ML 0.18
ns
7.19 0.002
ns
, Not signiﬁcant.
Fig.2.
Group(blackbars)effects(Re,regularstudents,Ph,physicaleducationstudents,Gy,studentsspecializedingymnastics)andconditioneffects(standingwitheyesopen,EO, eyes closed, EC, and on surface, FO) (grey bars) of mediolateral (ML) and anteriorposterior (AP) variability (RMS), mean power frequency (MPF), long-range correlations(
a
), regularity (SEn) and local stability (
l
max
). Error bars indicate standard errors.
C.J.C. Lamoth et al./Gait & Posture xxx (2009) xxx–xxx
4
G Model
GAIPOS-2711; No of Pages 6
Please cite this article in press as: Lamoth CJC, et al. Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics. Gait Posture (2009), doi:10.1016/j.gaitpost.2008.12.006
differentiate between three healthy young populations thatdiffered in athletic skill level. To this end, we examined thestructure of sway ﬂuctuations during quiet tandem stance undermanipulation of vision and proprioception. We expected thatcondition effects and group differences would be apparent in thestructure of variability of ML and AP accelerations. That is, weexpected that changes in variability would be accompanied bychanges in the regularity of the time-series and local stabilitymeasures, reﬂecting differences in postural efﬁciency. Further-more, we expected that accelerations in ML and AP of the regularstudent group would be characterized by greater variability(RMS), less regularity, and smaller local stability in the morechallenging conditions, i.e., standing with eyes closed and onfoam, than in the other two athletic groups. In addition, weexpected the gymnastic group to have smaller RMS values, lowerregularityandhigherlocalstabilitywhenstandingonanunstablefoam surface than the other two groups.Overall tandem stance with eyes closed and standing onfoam increased the magnitude of the ﬂuctuations (RMS) of theaccelerationsinbothAP andMLdirection,butmanipulatingvisionor proprioception had no differential effect on the accelerations,i.e.,changingproprioceptiveinformationdidnotresultindifferentRMS values than removing vision. Similarly, MPF of AP accelera-tions was greater with EC, but lower when standing on foam thanwith eyes closed. In-depth analyses revealed that the variabilityof the ML and AP acceleration time-series was not just randomnoise but contained deterministic components. For all outcomemeasures the estimates of the surrogate data were signiﬁcantlydifferent from those of the srcinal data. Long-range correlationswere present in all time-series, both with and without vision. InAP direction these correlations decreased, as did regularity (SEnincreased), whereas local stability increased (i.e.,
l
max
decreased).However, the priopiocepsis manipulation, i.e., standing on anunstable surface, had an opposite effect; long-range correlationsincreased while regularity and local stability decreased. For ML only long-range correlations were signiﬁcantly smaller than forstandingonfoam.TheseﬁndingsarelargelyinlinewiththeresultsofCOPdatafromthestudyofSchmitetal.[28]andRileyetal.[22],
who applied recurrence quantiﬁcation analysis to assess COPdynamics, and reported a trend toward increased regularity of theCOP time-series as balance conditions became more challenging.However, they reported no differential effects for vision andproprioception manipulations.The differential effects of condition for ML and AP accelerationare probably related to the task requirements of tandem stance.Tandem stance especially challenges ML movement as its base of support is very small and constrains postural adaptations.Consequently, condition effects were most prominent in the APdirection. In addition, tandem stance may require increasedcognitive involvement in postural control, thereby affecting swayregularityandlocalstability[8],whichwouldbeﬁtRoerdinketal.’s[24] suggestion that posture is controlled predominantly in thedirection of the largest sway.Although the three subject groups differed in athletic skill levelonly, consistent signiﬁcant group differences were found, withmoderate to very large effect sizes. As gymnastic skill levelincreased, acceleration variability decreased, regularity decreased,andlocalstabilityincreased,indicatingamoreautomaticandmoreefﬁcient postural control, in line with our expectations and thetheoretical notion of a continuum of postural variability. Sincegymnasts are more experienced in balance tasks such as landingand standing still on compliant foam mats, and since standing ondifferent kinds of surfaces is often part of their training, weexpected to ﬁnd signiﬁcant group by condition interaction effects.However, the unexpected ﬁnding that the observed groupdifferences did not depend on vision or proprioception (i.e, nosigniﬁcant group by condition interactions were found) suggestedthat expert gymnastics exhibited differences in the underlyingorganization of postural control which, at least for the presenttask, appear to be independent of the speciﬁc form of sensoryinformation used.In the present study body sway was measured using trunkaccelerometrywhichhastheadvantagethatareferencepointintheproximity of the COM is used. The results clearly show that acombination of accelerometry and off-line stochastic-dynamicalanalysis can discriminate the sway patterns produced by differentsubject groups. Although no clinical conclusions can be drawnfrom this non-clinical ﬁnding, it does suggest that quantifyingthe dynamics structure of postural sway using an ambulantaccelerometer bases system may provide a useful diagnostic tool.Notwithstanding the immediate advantages of accelerometricsystemsforclinicalpurposes(i.e.,compactandeasy-to-use,minimalawarenessofthemeasuringprocessonthepartofthesubject),theyalsohaveseveraldrawbackssuchasthe,theneedtopreprocessthedata(tocorrectfortiltandtoeliminatenoise),andthetranslationtoclinically understandable outcome measures. These processes arebeing automatedandsimpliﬁedforclinicaluse (e.g.,DynaPort
1
Mi-niMod, McRoberts BV, The Hague, the Netherlands). Future studiesshould reveal the degree to which this promise holds in pursuingspeciﬁc clinical research questions and objectives.
Conﬂict of interest statement
None of the authors has any ﬁnancial and/or personalrelationships with other people or organisations that could haveinappropriately inﬂuence (bias) the work.
Fig.3.
Randomizationeffects forlong-range correlations (
a
), local stability (
l
max
) andregularity (SEn). Asterisks indicate signiﬁcanteffects. ORIG,srcinalacceleration data;TIME, shufﬂed (time) randomized data; PHASE, phase-randomized data. Note that for the shufﬂed data
a
= 0.5, and the SampEn increases signiﬁcantly. For
l
max
no shufﬂedcounterparts can be calculated. Error bars indicate standard errors.
C.J.C. Lamoth et al./Gait & Posture xxx (2009) xxx–xxx
5
G Model
GAIPOS-2711; No of Pages 6
Please cite this article in press as: Lamoth CJC, et al. Athletic skill level is reﬂected in body sway: A test case for accelometry incombination with stochastic dynamics. Gait Posture (2009), doi:10.1016/j.gaitpost.2008.12.006

Related Search

We Need Your Support

Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks