Effects of a cognitive dual task on variability and local dynamic stability in sustained repetitive arm movements using principal component analysis: a pilot study

Effects of a cognitive dual task on variability and local dynamic stability in sustained... In many daily jobs, repetitive arm movements are performed for extended periods of time under continuous cognitive demands. Even highly monotonous tasks exhibit an inherent motor variability and subtle fluctuations in movement stability. Variability and stability are different aspects of system dynamics, whose magnitude may be further affected by a cognitive load. Thus, the aim of the study was to explore and compare the effects of a cognitive dual task on the variability and local dynamic stability in a repetitive bimanual task. Thirteen healthy volunteers performed the repetitive motor task with and without a concurrent cognitive task of counting aloud backwards in multiples of three. Upper-body 3D kinematics were col- lected and postural reconfigurations—the variability related to the volunteer’s postural change—were determined through a principal component analysis-based procedure. Subsequently, the most salient component was selected for the analysis of (1) cycle-to-cycle spatial and temporal variability, and (2) local dynamic stability as reflected by the largest Lyapunov exponent. Finally, end-point variability was evaluated as a control measure. The dual cognitive task proved to increase the temporal variability and reduce the local dynamic stability, marginally decrease endpoint variability, and substantially lower the inci- dence of postural reconfigurations. Particularly, the latter effect is considered to be relevant for the prevention of work-related musculoskeletal disorders since reduced variability in sustained repetitive tasks might increase the risk of overuse injuries. Keywords Dual task · Largest Lyapunov exponent · Movement variability · Musculoskeletal disorders (MSDs) · Postural reconfigurations · Principal component analysis (PCA) Introduction movements. In fact, movement variability has been con- sidered a prerequisite for flexibility and adaptability, both Movement variability is a pervasive and fundamental aspect crucial to motor learning (see, e.g., Wolpert et al. 2001; Dha- of human performance. The redundancy of the motor sys- wale et al. 2017), which may have implications for the pre- tem allows for the use of multiple strategies to perform any vention of overuse injuries (Hamill et al. 2012; Stergiou and given task. Therefore, even highly monotonous tasks exhibit Decker 2011). It has been suggested that one way to prevent substantial variation over repetitions (Bernstein 1967). This overuse injuries or pain is to regularly alter the movement inherent motor variability, which can manifest itself both pattern in the execution of the repetitive task, thereby avoid- in movements and in postures (Srinivasan and Mathiassen ing an overload of the same soft tissues (Bartlett et al. 2007). 2012), may be an important index of healthy and functional This hypothesis is of particular relevance in occupational contexts (Srinivasan and Mathiassen 2012; Madeleine et al. 2008; Fuller et al. 2011; Côté et al. 2005) for the prevention of work-related musculoskeletal disorders. * Alessia Longo a.longo@donders.ru.nl Characterizing movement variability remains an impor- tant challenge, since several different methods have been Donders Institute for Brain, Cognition and Behaviour, used to quantify movement variations, which not necessar- Radboud University, P.O. Box 9104, 6500 HE Nijmegen, ily have the same meaning (Van Emmerik et al. 2016). A The Netherlands 2 traditional way to quantify movement variability is to use Department of Sport Science, University of Innsbruck, discrete movement variables such as the standard deviation Fürstenweg 185, 6020 Innsbruck, Austria Vol.:(0123456789) 1 3 1612 Experimental Brain Research (2018) 236:1611–1619 of movement amplitudes, i.e., spatial variability (Cignetti are less stable (Woollacott and Shumway-Cook 2002; Mag- et  al. 2009) or cycle durations, i.e., temporal variability nani et  al. 2017). The effects of dual-task paradigms on (Danion et al. 2014). As opposed to traditional linear meas- movement variability, however, have mostly been contro- ures, the dynamical system theory (Kelso 1995) takes into versial (Beurskens and Bock 2013; Beauchet et al. 2005). account both spatial and temporal aspects of the movement In the current study, we designed a repetitive bimanual and emphasizes notions such as stability and critical fluctua - task with no postural constraints, which resembles real work- tions to capture essential features of movements (Harbourne related environments. In the context of the just described and Stergiou 2009), shifting the focus from isolated joints views on movement variability and stability, we here pro- towards complex coordinated actions (Bartlett et al. 2007). pose an alternative analysis method. The approach consists The basic assumption of the dynamical system theory is that of applying principal component analysis (PCA) on the any multisegmental biological system which shows coor- subjects’ upper-body postural motion (Daffertshofer et al. dinated motor behavior by activity at the level of muscles, 2004; Federolf et al. 2012) to isolate variability related to joints, and limbs, will find stable macroscopic coordination postural reconfigurations, i.e., intermittent and incidental patterns by means of self-organization due to the intrinsic changes in posture. We consider such postural configura- dynamics of the interactions at the microscopic level of its tions as non-linear transitions between two different (pos- segments. The term stability in this context refers to the tural) coordination patterns. Our variability and stability capacity of the system to counteract perturbations (Ding- analyses subsequently were directed to the most salient well and Marin 2006). The largest Lyapunov exponent is component following the PCA, which, in a sense, was not a nonlinear measure used to determine the local aspects of ‘contaminated’ by the non-linear, postural reconfigurations. stability (Segal et al. 2008; Dingwell et al. 2001; Hak et al. Here, different variables were calculated (1) cycle-to-cycle 2013). Local dynamic stability refers to the sensitivity of a spatial and temporal variability and (2) local dynamic sta- system to small, intrinsic perturbations, and should not be bility as reflected by the largest Lyapunov exponent. As a confused with global stability. In fact, local uc fl tuations need control measure, endpoint variability was also assessed. In to be attenuated to maintain global stability (Van Emmerik our view, exploring different methods to quantify movement et al. 2016). variations at the level of the whole upper body may increase Variability and stability, although related, represent dif- our understanding of the role of stability versus variability ferent concepts. Their exact relationship is not clear yet. On in sustained cyclical motion. the one side, an increase in movement variability is con- In summary, the goal of the current study was to inves- sidered a source of behavioral change in the system which tigate the effects of a concurrent cognitive task on differ - signifies growing instability that may lead to a coordina- ent types of movement variability i.e., endpoint variability, tion shift to a different stable coordination pattern. On the postural reconfiguration, cycle-to-cycle variability, and on other side, some behaviors which seem to be stable, may the local dynamic stability, i.e., largest Lyapunov exponent, paradoxically show quite some variability (Dingwell and in a sustained repetitive upper-extremity motor task. In line Marin 2006). Thus, it seems that variability does not always with the view that tasks that demand more resources are decrease when people get into or refine a stable behavioral less stable, we hypothesized that the local dynamic stability, state. In certain conditions, variability may actually increase. would decrease in the dual-task condition. With respect to This contradictory relationship is noticeable when observ- movement variability, no particular effects were predicted ing the rich behavioral repertoire of elite sport players or because of contradictory or absence of earlier findings. Our expert musicians (Harbourne and Stergiou 2009; Glazier pilot study was conducted to increase our understanding on et al. 2003). how a cognitive load may contribute to an increased risk for An effective way to manipulate variability and stability in work-related overuse injuries, a topic to which we will return a cyclical motor task is adding a secondary cognitive task. in the “Discussion” section. This method is of particular interest in the context of the risks for work-related MSDs since cognitive demands are a relevant occupational factor which has been shown to affect Methods sustained repetitive movements (Srinivasan and Mathias- sen 2012; Bloemsaat et al. 2005). The underlying theory Participants of studies on dual-tasking is that resources are limited, and they have to be shared between a cognitive and a motor task, Thirteen right-handed healthy subjects (9 female, 4 male; consequently performance will suffer (Plummer and Eskes 25.46 ± 3.46 years) volunteered for this study. No participant 2015). In dual-task paradigms, local dynamical stability reported pain or history of injuries in neck, shoulder and might decrease, in terms of limited resources, since more arm regions. All participants gave informed written consent difficult tasks demand more resources and as a consequence 1 3 Experimental Brain Research (2018) 236:1611–1619 1613 and the study was approved by an institutionalized ethics the chair was then maintained in all trials. All participants review board. performed a warming-up trial for a minimum of 30 s or until they felt comfortable with the task. Procedure Equipment The protocol consisted of two trials of 5 min each, whose order was counterbalanced between participants. In one A custom-made script in Python 2.7 2010 (Python Soft- trial, a motor task was performed solo (M) and in the other ware Foundation, Beaverton, OR, USA) was implemented a motor task was performed in combination with a cogni- to present the stimuli and record the endpoint position on tive task (M + C). In the cognitive task, participants counted a 27′′ touch screen (1920 × 1080 resolution; ProLiteb Iiy- aloud backwards in multiples of three. In the motor task, ama, Iiyama Corporation, Tokyo, Japan). Four targets of participants performed a sustained repetitive task on a multi- 27 mm in diameter were presented, with a between-targets touch screen, tapping two pairs of visually presented targets distance of 125 mm in anterior direction and a distance of with both hands simultaneously and in-phase (Fig. 1). Par- 155 mm between the targets of the two hands. Xsens MVN ticipants could perform the task freely (without specified BIOMECH motion capture suit (Xsens technologies BV, rhythm or posture), with the only requirement of touching Enschede, The Netherlands) was used to record upper body the targets as fast and as accurate as possible. The motor task kinematics at 60 Hz. Eleven sensor units were placed on the shares common features with the bimanual Fitts’ task used head, sternum, pelvis, shoulders, upper arms, forearms and in previous studies (Longo and Meulenbroek 2018; Shea hands following the recommendations by Xsens. Anatomi- et al. 2012; Amazeen et al. 2005), however, no specific task cal measurements and calibrations were performed accord- variations as regards movement amplitude and target width, ing to the procedures provided by Xsens. Data acquisition were applied in the current study to enhance its monotony. was done via the accompanying software (MVN Studio 4.2, Before starting the measurements, participants were asked Xsens technologies BV, Enschede, Netherlands) which cal- to adjust the chair height and distance to the touch screen culates the kinematic data. Thirty joint angles: the 3D angle to find the most comfortable position. The configuration of configurations of wrist (2: left and right), elbow (2), shoul- der (2) and four column angles: C1–Head, T1–C7, T9–T8, L5–S1, were considered for further analysis. All joint angles were expressed in local coordinate systems following the guidelines of the International Society of Biomechanics (Wu et al. 2005). Data analysis Touch screen data (XY coordinates) of the fingertip posi - tions realized during the 5-min task were used to quantify the endpoint variability. We used the standard measure of variable error (VE), which was defined as the mean distance of all movement endpoints to the mean endpoint (Gordon et  al. 1994). For further calculations, we determined for each participant the mean of VE for all four targets (VE ). One subject was excluded from this analysis since data were missing. Xsens data were used to investigate other types of vari- ability and local dynamic stability, which were the primary interest in the present study. For this purpose, a PCA was applied using the 30 joint angles as 30-dimensional input vectors. Prior to data analysis, the first 5 s were excluded from the raw dataset, to avoid analyzing settling-in behav- ior. For each trial, every angle vector was normalized by subtracting the trial-mean. Then, a single input matrix was created with the normalized vectors as columns and the data of all subjects and both conditions (M, M + C) concat- enated vertically. Finally, a single PCA was calculated on Fig. 1 Experimental setup 1 3 1614 Experimental Brain Research (2018) 236:1611–1619 this combined input matrix to facilitate direct comparisons series from its trend was lower than half its total average between participants (Federolf 2016; Gløersen et al. 2017). (i.e., pause within the repetitive task), or if it exceeded two The first three principal components (PCs) were considered times its total average (e.g., unusual movement). The slope for the analysis of postural reconfigurations which reflect of the trends underlying the residual time periods were used the movement variability related to the volunteer’s postural to delineate transitions, non-stationary, and quasi-stationary changes ( see also Longo et al. 2018). The first principal phases, respectively. Specifically, transitions were defined if component was further examined for the analysis of cycle- the absolute value of the slope for a minimum of 100 sam- to-cycle variability and local dynamic stability in repetitive ples exceeded a threshold of 0.1 and non-stationary phases cycles. All calculations in the current study were imple- if the absolute value of the slope for a minimum of 300 sam- mented in Matlab R2015a (MathWorks Inc., Natic, MA, ples exceeded a threshold of 0.02. Thresholds and number USA). of samples used were specific for our setup and best identi- fied the four phases. Time periods that were not allocated to Postural reconfigurations any of the former phases were marked as quasi-stationary phases. If a criterion for any of the phases was met in one Changes in the postural configuration (Fig.  2) were deter- PC, then this period was delineated accordingly in all PC mined by first defining the trends of the first three PCs (black time series (Fig. 2). For further comparisons, the cumulated lines) through a low pass filter (Butterworth filter; cut-off duration per minute of each phase (D) was calculated. Thus, 0.1 Hz). Then, the trends were used to classify four phases: for each condition (M, M + C), four dependent variables were events—interruptions or unusual movements during the defined: D (events), D (transitions), D (non-stationary e t ns execution of the task; transitions—rapid changes from a phases), and D (quasi-stationary phases). qs postural configuration to another; non-stationary phases— gradual changes between postural configurations; quasi-sta- Cycle‑to‑cycle variability tionary phases—unchanging postural configurations. In par - ticular, events were defined by subtracting each trend from Thirty consecutive cycles were selected in the PC1 time its PC; an event was marked if the deviation of the PC time series in a quasi-stationary phase (Fig.  3a). The cycles M+C -20 -40 050 100 150 200 250 050100 150200 250 -20 -40 -20 -40 050100 150200 250 050 100 150200 250 time [s] Fig. 2 Representative dataset of a 5  min trial of the motor (M) and stationary phases (cyan), non-stationary phases (green), and transi- the motor + cognitive (M + C) trial of one arbitrarily selected volun- tions (red). The black line represents the low pass-filtered underlying teer: the first three PCs are shown. The tapping movement between trend two pairs of targets is printed as a colored line, respectively, quasi- 1 3 PC3 PC1 PC2 Experimental Brain Research (2018) 236:1611–1619 1615 selected corresponded to the first quasi-stationary phase exponential rate of separation of neighboring trajectories of of at least 30 cycles, i.e., longest consecutive cycles that the attractor. LyE was calculated by first constructing a state could be detected among all participants and both condi- space representation of the time series (Fig. 3b). The time tions in the relevant time periods. A cycle was defined delay (τ) was determined using the average mutual infor- as a back and forward movement, starting from the tar- mation (AMI; Fraser and Swinney 1986 )and the embed- gets closer to the body. The starting points of the cycles ding dimension (m) using a false nearest neighbor algorithm corresponded to the local maxima of PC1. Spatial (SD ) (Kantz 1994) Therefore, m = 2 and τ = 9 were selected. and temporal (SD ) variability were calculated on the 30 Finally, LyE values were calculated for the time series using selected cycles. SD was calculated by first interpolating Wolf’s algorithm (Wolf et al. 1985; Buzzi et al. 2003.) each cycle such that it was represented by 100 samples (i.e., expressed in percent). For each sample, the stand- Statistical analysis ard deviation between cycles was determined. Finally, the mean of the standard deviations over the whole cycle was To determine changes in the postural configuration, the calculated. SD was assessed as the standard deviation of cumulated duration per minute of each phase (D , D , D , T e t ns the movement duration between cycles. The mean of the D ) was compared between the two conditions (M, M + C). qs movement duration (T ) between the 30-selected cycles As the data were not normally distributed, a Wilcoxon was also assessed as a control measure. signed-rank test was used. Variables D , D , D are inde- e t ns pendent and were analyzed applying a Šidák correction for Local dynamic stability multiple comparisons, thus reducing the α-level for sta- tistical significance to α = 0.0174. For completeness, also The largest Lyapunov exponent (LyE) was calculated for variable D , which directly depends on the other variables qs the same 30 cycles selected in the PC1-time series. LyE is (D  = 60 s − [D  + D  + D ]) was analyzed; also applying qs e t ns a measure of local dynamic stability, which quantifies the the corrected α-threshold of α = 0.0174. For the analysis of M+C (a) -20 -40 050100 150200 250 050 100 150 200 250 time [s] time [s] (b) LyE = LyE =3.42 LyE =1.97 Fig. 3 a Representation of PC1 of the motor (M) and the motor + cog- ysis of cycle-to-cycle variability. b State space representation of 30 nitive (M + C) trial of one arbitrary selected subject. The enlargement cycles of the same representative subject for the analysis of the larg- shows 30 cycles selected in the quasi-stationary phases for the anal- est Lyapunov exponent 1 3 PC1 1616 Experimental Brain Research (2018) 236:1611–1619 cycle-to-cycle variability, local dynamic stability and end- to PC1 and were largely affected by postural reconfigurations point variability the data were normally distributed, there- of the subjects. fore, a paired-samples t test was used to compare SD SD , C, T T , LyE, and VE for both conditions. Here, the α-level for Motor task versus motor + cognitive task m m statistical significance was set to α = 0.05. Statistical analy- ses were performed using SPSS Version 22 (IBM, Chicago, With respect to endpoint variability, a statistical trend IL, USA). was observed in VE which decreased in the M + C [9.34 (± 3.22) mm] compared to the M [11.01 (± 3.78) mm] trials [t(11) = 1.83, P = 0.095, d = 0.48]. Postural reconfigurations Results (Fig. 4a) revealed a significant main effect in D which was qs higher in the M + C than in the M trial (Z = 2.43, P = 0.015, Results of the principal component analysis r = 0.67). A statistical trend was found in D (Z = 2.02, P = 0.043, r = 0.56), indicating more frequent changes in the Principal components 1–3 represented 44.9, 16.9, and 9.9% M than in the M + C trial. No significant differences between of the overall variance in the kinematic data, respectively. conditions were found in D (Z = 1.73, P = 0.084, r = 0.48) ns Figure 2 shows an example of the first three PC score time and in D (Z = 0.67, P = 0.5, r = 0.19). series of the M and M + C trial for one selected subject. The Cycle-to-cycle variability (Fig.  4b) revealed a signifi- first principal component (PC1) represented the movement cant main effect between conditions in SD [t(12) = 2.39, component containing the largest variance and, in the cur- P = 0.036, d = 0.75] indicating higher temporal variability rent study, PC1 was dominated by the cyclic movement pat- in the M + C than in the M trial. However, no significant tern of the task. PC2 and PC3 represent variance orthogonal differences between conditions were observed in SD Changes in the postural configuration (a) * 2 25 25 20 20 1.5 15 15 10 10 0.5 5 5 0 0 0 0 MM+C MM+C MM+C MM+C Cycle-to-cycle variability Local dynamic stability (b) (c) 0.08 0.06 0.04 3 0.02 M M+C M M+C MM+C Fig. 4 a Box plots of cumulative duration per minute of events (D ; (SD ) and temporal variability (SD ) for M and M + C; c box plot of e C T magenta), transitions (D ; red), non-stationary phases (D ; green) the largest Lyapunov exponent (LyE) for M and M + C. Significant t ns and quasi-stationary phases (D ; cyan)for the motor (M) and the between condition effects are indicated by an asterisk qs motor + cognitive (M + C) trial; b box plots of spatial variability 1 3 D [s] SD D [s] SD [s] D [s] ns LyE D [s] qs Experimental Brain Research (2018) 236:1611–1619 1617 [t(12) = 0.45, P = 0.664, d = 0.16] and in T [t(12) = 1.24, due to a concurrent cognitive task, as we observed in the P = 0.239, d = 0.33]. Local dynamical stability (Fig.  4c) current study, local stability may be reduced. A dynamical decreased in the M + C as compared to the M trial as movement system can try to attenuate local fluctuations and reflected by an increase of  LyE [t (12) = 2.36, P = 0.036, maintain a stable coordination pattern by adopting another d = 0.6]. functional solution or coordination mode that suits the dual- task constraints better. The result of this process may be that the system is constrained at the joint level thus reducing the Discussion incidence of postural reconfigurations. A novelty of the present study is the application of PCA The current pilot study explored the effects of a concurrent for the assessment of different types of variability and local- cognitive task on different types of movement variability, dynamic stability. Using this approach, we moved away from i.e., endpoint variability, postural reconfiguration, cycle- quantyfing the variability of isolated joints by a limited to-cycle variability, and on the local dynamic stability, i.e., number of pre-selected kinematic variables, and instead, largest Lyapunov exponent, in a sustained repetitive upper- moved towards metrics such as postural reconfigurations of extremity motor task. In agreement with our hypothesis and the whole upper body which allowed us to capture complex the view that tasks that demand more resources are less sta- multijoint coordination and thus provide a fuller account of ble, the local dynamic stability decreased under dual-task multijoint cyclical movements while coping with a cogni- conditions (Fig. 4c). However, the effects of the secondary tive load. Further, we attempted to better understand what cognitive load on different types of movement variability distinct parameters measuring variability and stability reflect revealed contrasting results as compared to earlier studies in sustained upper-extremity motion. Our results show that that used different variability measures (Beauchet et  al. LyE and temporal variability reflect unwanted fluctuations 2005; Hollman et al. 2007). Temporal variability increased in performance due to reduced control with an increase in (Fig. 4b), suggesting that the cognitive task caused interfer- task difficulty.The incidence of postural reconfigurations, ence due to the competition for attentional resources neces- however, reflects a potential beneficial variability due to the sary for the motor task. The increase of temporal variability dynamics of the human movement system. Another benefit with an additional cognitive load is in line with dual-task of distinguishing between different types of variability by interference effects reported earlier (Dubost et  al. 2006; means of PCA is that the LyE can then be calculated on Beauchet et al. 2005). Spatial variability was not affected quasi-stationary phases. Stationarity of the underling time by the counting task and the endpoint variability marginally series is a prerequisite for this calculation, but in human decreased. Simultaneously, the incidence of postural recon- movement studies this stationarity is often difficult to define. gu fi rations signica fi ntly decreased in the dual-task condition However, one limitation of this approach was that the num- (Fig.4a), indicating that participants adopted fixed postures ber of consecutive cycles needed for the calculation of LyE for longer periods of time. Since motor variability has been was limited by the occurrence of postural reconfigurations, purported as beneficial for avoiding overuse injuries and nonetheless the length selected is considered adequate for pain (Bartlett et al. 2007; Srinivasan and Mathiassen 2012), the analysis (cf. Wolf et al. 1985). Further, due to the novelty the decrease in postural readjustments due to dual tasking of the current approach and the low sample size, our findings may constitute a risk factor for MSDs. need to be taken with caution. The postural readjustment results may be interpreted In conclusion, the current findings suggest that under from the viewpoint of dynamical systems theory as follows. cognitive demands, the temporal variability and dynamic Generally, in dual-task paradigms, the challenge for a motor instability of cyclical arm movements increase. Simultane- system performing movements and a cognitive task is to ously, at the postural level, cognitive loads led to a decreased adapt to the secondary task demands without reducing the incidence of postural reconfigurations. Particularly, for the quality of movement performance. The main purpose of a prevention of MSDs, this reduced postural variability should dynamical system then is to reach or maintain global sta- be carefully monitored since postural reconfigurations may bility. Goal-directed actions are supported by reducing the play a role in the prevention of overuse injuries. number of biomechanical degrees of freedom of the motor Acknowledgements This study was n fi ancially supported by the Euro - system through the formation of functional synergies afford- pean Union FP7 Marie Curie IDP Grant (FP7-PEOPLE-2013-ITN ing preferred and stable coordination patterns. However, a ‘HealthPAC’, Grant 604063-IDP). We wish to acknowledge the sup- stable system does permit flexible and adaptive motor behav - port of Henk Luinge and Matteo Giuberti (Xsens technologies BV, Enschede, The Netherlands) for providing the sensors and helping with ior, encouraging free exploration of coordination changes the preparation of the experiment. to be able to acquire different stable motor solutions over time, a mechanism known to enhance motor learning (Gla- zier et  al. 2003). Now, if the motor system is perturbed 1 3 1618 Experimental Brain Research (2018) 236:1611–1619 Dhawale AK, Smith MA, Olveczky BP (2017) The role of variability Compliance with ethical standards in motor learning. Annu Rev Neurosci 40:479–498. https ://doi. org/10.1146/annur ev-neuro -07211 6 Conflict of interest The authors have no conflicts of interest. Dingwell JB, Marin LC (2006) Kinematic variability and local dynamic stability of upper body motions when walking at differ - Ethical standards The study was approved by an institutionalized ethics ent speeds. J Biomech 39(3):444–452. https ://doi.org/10.1016/j. review board in accordance with the ethical standards laid down in the jbiom ech.2004.12.014 1964 Declaration of Helsinki. Dingwell JB, Cusumano JP, Cavanagh PR, Sternad D (2001) Local dynamic stability versus kinematic variability of continuous Informed consent All participants gave informed written consent prior overground and treadmill walking. J Biomech Eng 123(1):27–32. to their inclusion in the study. http://www.ncbi.nlm.nih.gov/pubme d/11277 298 Dubost V, Kressig RW, Gonthier R, Herrmann FR, Aminian K, Najafi B, Beauchet O (2006) Relationships between dual-task Open Access This article is distributed under the terms of the Crea- related changes in stride velocity and stride time variability in tive Commons Attribution 4.0 International License (http://creat iveco healthy older adults. Hum Mov Sci 25(3):372–382. https ://doi. mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- org/10.1016/j.humov .2006.03.004 tion, and reproduction in any medium, provided you give appropriate Federolf PA (2016) A novel approach to study human posture control: credit to the original author(s) and the source, provide a link to the “Principal movements” obtained from a principal component Creative Commons license, and indicate if changes were made. analysis of kinematic marker data. J Biomech 49(3):364–370. https ://doi.org/10.1016/j.jbiom ech.2015.12.030 Federolf P, Tecante K, Nigg B (2012) A holistic approach to study the temporal variability in gait. J Biomech 45(7):1127–1132. https :// References doi.org/10.1016/j.jbiom ech.2012.02.008 Fraser AM, Swinney HL (1986) Independent coordinates for strange Amazeen EL, Ringenbach SD, Amazeen PG (2005) The effects of attractors from mutual information. Phys Rev A 33(2):1134–1140. attention and handedness on coordination dynamics in a biman- https ://doi.org/10.1103/PhysR evA.33.1134 ual Fitts’ law task. Exp Brain Res 164(4):484–499. https ://doi. Fuller JR, Fung J, Côté JN (2011) Time-dependent adaptations to org/10.1007/s0022 1-005-2269-y posture and movement characteristics during the development of Bartlett R, Wheat J, Robins M (2007) Is movement variability impor- repetitive reaching induced fatigue. Exp Brain Res 211(1):133– tant for sports biomechanists? Sports Biomech 6(2):224–243. 143. https ://doi.org/10.1007/s0022 1-011-2661-8 https ://doi.org/10.1080/14763 14070 13229 94 Glazier PS, Davids K, Bartlett RM (2003) Dynamical systems theory: a Beauchet O, Dubost V, Herrmann FR, Kressig RW (2005) relevant framework for performance-oriented sports biomechanics Stride-to-stride variability while backward counting among research. Sportscience. http://shura .shu.ac.uk/13427 /. Accessed healthy young adults. J NeuroEng Rehabil. https ://doi. 2 Feb 2018 org/10.1186/1743-0003-2-26 Gløersen Ø, Myklebust H, Hallén J, Federolf P (2017) Technique anal- Bernstein N (1967) The co-ordination and regulation of movements. ysis in elite athletes using principal component analysis. J Sports Pergamon Press, Oxford Sci. https ://doi.org/10.1080/02640 414.2017.12988 26 Beurskens R, Bock O (2013) Does the walking task matter? Influ- Gordon J, Ghilardi MF, Cooper SE, Ghez C (1994) Accuracy of pla- ence of different walking conditions on dual-task performances nar reaching movements I. Independence of direction and extent in young and older persons. Hum Mov Sci 32:1456–1466. https variability. Exp Brain Res 99(1):112–130. https://doi.or g/10.1007/ ://doi.org/10.1016/j.humov .2013.07.013 BF002 41416 Bloemsaat JG, Meulenbroek RGJ, Van Galen GP (2005) Differential Hak L, Houdijk H, Beek PJ, Van Dieën JH (2013) Steps to take to effects of mental load on proximal and distal arm muscle activ - enhance gait stability: the effect of stride frequency, stride length, ity. Exp Brain Res 167(4):622–634. https://doi.or g/10.1007/s0022 and walking speed on local dynamic stability and margins of 1-005-0066-2 stability. PloS One 8(12):e82842. https ://doi.org/10.1371/journ Buzzi UH, Stergiou N, Kurz MJ, Hageman PA, Heidel J (2003) Non- al.pone.00828 42 linear dynamics indicates aging affects variability during gait. Hamill J, Palmer C, Van Emmerik REA (2012) Coordinative vari- Clin Biomech 18(5):435–443. https ://doi.or g/10.1016/S0268 ability and overuse injury. SMARTT 4(1):45. https ://doi. -0033(03)00029 -9 org/10.1186/1758-2555-4-45 Cignetti F, Schena F, Rouard A (2009) Effects of fatigue on inter-cycle Harbourne RT, Stergiou N (2009) Movement variability and the use variability in cross-country skiing. J Biomech 42(10):1452–1459. of nonlinear tools: principles to guide physical therapist practice. https ://doi.org/10.1016/j.jbiom ech.2009.04.012 Phys Ther 89(3):267–282. https ://doi.org/10.2522/ptj.20080 130 Côté JN, Raymond D, Mathieu PA, Feldman AG, Levin MF (2005) Hollman JH, Kovash FM, Kubik JJ, Linbo RA (2007) Age-related Differences in multi-joint kinematic patterns of repetitive ham- differences in spatiotemporal markers of gait stability dur- mering in healthy, fatigued and shoulder-injured individuals. Clin ing dual task walking. Gait Posture 26(1):113–119. https ://doi. Biomech 20(6):581–590. https ://doi.or g/10.1016/j.clinb iomec org/10.1016/j.gaitp ost.2006.08.005 h.2005.02.012 Kantz H (1994) A robust method to estimate the maximal Lyapunov Daffertshofer A, Lamoth CJC, Meijer OG, Beek PJ (2004) PCA in exponent of a time series. Phys Lett A 185(1):77–87. https ://doi. studying coordination and variability: a tutorial. Clin Bio- org/10.1016/0375-9601(94)90991 -1 mech 19(4):415–428. ht tp s : //d oi .o r g/ 10 .1 01 6/ j.c l inb i om ec Kelso JAS (1995) Dynamic patterns: the self-organization of brain and h.2004.01.005 behavior. The MIT Press, Cambridge Danion F, Bongers RM, Bootsma RJ (2014) The trade-off between Longo A, Meulenbroek R (2018) Precision-dependent changes in spatial and temporal variabilities in reciprocal upper-limb aiming motor variability during sustained bimanual reaching. Motor movements of different durations. PloS One 9(5):e97447. https: // Control 22:28–44. https ://doi.org/10.1123/mc.2016-0013 doi.org/10.1371/journ al.pone.00974 47 Longo A, Meulenbroek R, Haid T, Federolf P (2018) Postural reconfig- uration and cycle-to-cycle variability in patients with work-related 1 3 Experimental Brain Research (2018) 236:1611–1619 1619 musculoskeletal disorders compared to healthy controls and in Stergiou N, Decker LM (2011) Human movement variability, nonlin- relation to pain emerging during a repetitive movement task. ear dynamics, and pathology: is there a connection? Hum Mov Clin Biomech 54:103–110. https ://doi.org/10.1016/j.clinb iomec Sci 30(5):869–888. https://doi.or g/10.1016/j.humov.2011.06.002 h.2018.03.004 Van Emmerik REA, Ducharme SW, Amado AC, Hamill J (2016) Com- Madeleine P, Mathiassen SE, Arendt-Nielsen L (2008) Changes in paring dynamical systems concepts and techniques for biome- the degree of motor variability associated with experimental chanical analysis. J Sport Health Sci. https ://doi.org/10.1016/j. and chronic neck-shoulder pain during a standardised repeti- jshs.2016.01.013 tive arm movement. Exp Brain Res 185(4):689–698. https ://doi. Wolf A, Swift JB, Swinney HL, Vastano JA (1985) Determining Lya- org/10.1007/s0022 1-007-1199-2 punov exponents from a time series. Phys D Nonlinear Phenom Magnani RM, Lehnen GC, Rodrigues FB, de Sá e Souza GS, de 16(3):285–317. https ://doi.org/10.1016/0167-2789(85)90011 -9 Oliveira Andrade A, Vieira MF (2017) Local dynamic sta- Wolpert DM, Ghahramani Z, Flanagan JR (2001) Perspectives and bility and gait variability during attentional tasks in young problems in motor learning. Trends Cognit Sci 5(11):487–494. adults. Gait Posture 55:105–108. https ://doi.org/10.1016/j.gaitp https ://doi.org/10.1016/S1364 -6613(00)01773 -3 ost.2017.04.019 Woollacott M, Shumway-Cook A (2002) Attention and the control Plummer P, Eskes G (2015) Measuring treatment effects on dual-task of posture and gait: a review of an emerging area of research. performance: a framework for research and clinical practice. Front Gait Posture 16(1):1–14. https ://doi.or g/10.1016/S0966 Hum Neurosci 9:225. https://doi.or g/10.3389/fnhum.2015.00225 -6362(01)00156 -4 Segal AD, Orendurff MS, Czerniecki JM, Shofer JB, Klute GK Wu G, van der Helm FCT, Veeger HEJD, Makhsous M, Van Roy P, (2008) Local dynamic stability in turning and straight-line gait. Anglin C, International Society of Biomechanics, et al (2005) J Biomech 41(7):1486–1493. https ://doi.org/10.1016/J.JBIOM ISB recommendation on definitions of joint coordinate systems ECH.2008.02.012 of various joints for the reporting of human joint motion—part II: Shea CH, Boyle J, Kovacs AJ (2012) Bimanual Fitts’ tasks: Kelso, Sou- shoulder, elbow, wrist and hand. J Biomech 38(5):981–992. https thard, and Goodman, 1979 revisited. Exp Brain Res 216(1):113–://doi.org/10.1016/J.JBIOM ECH.2004.05.042 121. https ://doi.org/10.1007/s0022 1-011-2915-5 Srinivasan D, Mathiassen SE (2012) Motor variability in occupational health and performance. Clin Biomech. https: //doi.org/10.1016/j. clinb iomec h.2012.08.007 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Experimental Brain Research Springer Journals

Effects of a cognitive dual task on variability and local dynamic stability in sustained repetitive arm movements using principal component analysis: a pilot study

Free
9 pages
Loading next page...
 
/lp/springer_journal/effects-of-a-cognitive-dual-task-on-variability-and-local-dynamic-sth275lZHV
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Biomedicine; Neurosciences; Neurology
ISSN
0014-4819
eISSN
1432-1106
D.O.I.
10.1007/s00221-018-5241-3
Publisher site
See Article on Publisher Site

Abstract

In many daily jobs, repetitive arm movements are performed for extended periods of time under continuous cognitive demands. Even highly monotonous tasks exhibit an inherent motor variability and subtle fluctuations in movement stability. Variability and stability are different aspects of system dynamics, whose magnitude may be further affected by a cognitive load. Thus, the aim of the study was to explore and compare the effects of a cognitive dual task on the variability and local dynamic stability in a repetitive bimanual task. Thirteen healthy volunteers performed the repetitive motor task with and without a concurrent cognitive task of counting aloud backwards in multiples of three. Upper-body 3D kinematics were col- lected and postural reconfigurations—the variability related to the volunteer’s postural change—were determined through a principal component analysis-based procedure. Subsequently, the most salient component was selected for the analysis of (1) cycle-to-cycle spatial and temporal variability, and (2) local dynamic stability as reflected by the largest Lyapunov exponent. Finally, end-point variability was evaluated as a control measure. The dual cognitive task proved to increase the temporal variability and reduce the local dynamic stability, marginally decrease endpoint variability, and substantially lower the inci- dence of postural reconfigurations. Particularly, the latter effect is considered to be relevant for the prevention of work-related musculoskeletal disorders since reduced variability in sustained repetitive tasks might increase the risk of overuse injuries. Keywords Dual task · Largest Lyapunov exponent · Movement variability · Musculoskeletal disorders (MSDs) · Postural reconfigurations · Principal component analysis (PCA) Introduction movements. In fact, movement variability has been con- sidered a prerequisite for flexibility and adaptability, both Movement variability is a pervasive and fundamental aspect crucial to motor learning (see, e.g., Wolpert et al. 2001; Dha- of human performance. The redundancy of the motor sys- wale et al. 2017), which may have implications for the pre- tem allows for the use of multiple strategies to perform any vention of overuse injuries (Hamill et al. 2012; Stergiou and given task. Therefore, even highly monotonous tasks exhibit Decker 2011). It has been suggested that one way to prevent substantial variation over repetitions (Bernstein 1967). This overuse injuries or pain is to regularly alter the movement inherent motor variability, which can manifest itself both pattern in the execution of the repetitive task, thereby avoid- in movements and in postures (Srinivasan and Mathiassen ing an overload of the same soft tissues (Bartlett et al. 2007). 2012), may be an important index of healthy and functional This hypothesis is of particular relevance in occupational contexts (Srinivasan and Mathiassen 2012; Madeleine et al. 2008; Fuller et al. 2011; Côté et al. 2005) for the prevention of work-related musculoskeletal disorders. * Alessia Longo a.longo@donders.ru.nl Characterizing movement variability remains an impor- tant challenge, since several different methods have been Donders Institute for Brain, Cognition and Behaviour, used to quantify movement variations, which not necessar- Radboud University, P.O. Box 9104, 6500 HE Nijmegen, ily have the same meaning (Van Emmerik et al. 2016). A The Netherlands 2 traditional way to quantify movement variability is to use Department of Sport Science, University of Innsbruck, discrete movement variables such as the standard deviation Fürstenweg 185, 6020 Innsbruck, Austria Vol.:(0123456789) 1 3 1612 Experimental Brain Research (2018) 236:1611–1619 of movement amplitudes, i.e., spatial variability (Cignetti are less stable (Woollacott and Shumway-Cook 2002; Mag- et  al. 2009) or cycle durations, i.e., temporal variability nani et  al. 2017). The effects of dual-task paradigms on (Danion et al. 2014). As opposed to traditional linear meas- movement variability, however, have mostly been contro- ures, the dynamical system theory (Kelso 1995) takes into versial (Beurskens and Bock 2013; Beauchet et al. 2005). account both spatial and temporal aspects of the movement In the current study, we designed a repetitive bimanual and emphasizes notions such as stability and critical fluctua - task with no postural constraints, which resembles real work- tions to capture essential features of movements (Harbourne related environments. In the context of the just described and Stergiou 2009), shifting the focus from isolated joints views on movement variability and stability, we here pro- towards complex coordinated actions (Bartlett et al. 2007). pose an alternative analysis method. The approach consists The basic assumption of the dynamical system theory is that of applying principal component analysis (PCA) on the any multisegmental biological system which shows coor- subjects’ upper-body postural motion (Daffertshofer et al. dinated motor behavior by activity at the level of muscles, 2004; Federolf et al. 2012) to isolate variability related to joints, and limbs, will find stable macroscopic coordination postural reconfigurations, i.e., intermittent and incidental patterns by means of self-organization due to the intrinsic changes in posture. We consider such postural configura- dynamics of the interactions at the microscopic level of its tions as non-linear transitions between two different (pos- segments. The term stability in this context refers to the tural) coordination patterns. Our variability and stability capacity of the system to counteract perturbations (Ding- analyses subsequently were directed to the most salient well and Marin 2006). The largest Lyapunov exponent is component following the PCA, which, in a sense, was not a nonlinear measure used to determine the local aspects of ‘contaminated’ by the non-linear, postural reconfigurations. stability (Segal et al. 2008; Dingwell et al. 2001; Hak et al. Here, different variables were calculated (1) cycle-to-cycle 2013). Local dynamic stability refers to the sensitivity of a spatial and temporal variability and (2) local dynamic sta- system to small, intrinsic perturbations, and should not be bility as reflected by the largest Lyapunov exponent. As a confused with global stability. In fact, local uc fl tuations need control measure, endpoint variability was also assessed. In to be attenuated to maintain global stability (Van Emmerik our view, exploring different methods to quantify movement et al. 2016). variations at the level of the whole upper body may increase Variability and stability, although related, represent dif- our understanding of the role of stability versus variability ferent concepts. Their exact relationship is not clear yet. On in sustained cyclical motion. the one side, an increase in movement variability is con- In summary, the goal of the current study was to inves- sidered a source of behavioral change in the system which tigate the effects of a concurrent cognitive task on differ - signifies growing instability that may lead to a coordina- ent types of movement variability i.e., endpoint variability, tion shift to a different stable coordination pattern. On the postural reconfiguration, cycle-to-cycle variability, and on other side, some behaviors which seem to be stable, may the local dynamic stability, i.e., largest Lyapunov exponent, paradoxically show quite some variability (Dingwell and in a sustained repetitive upper-extremity motor task. In line Marin 2006). Thus, it seems that variability does not always with the view that tasks that demand more resources are decrease when people get into or refine a stable behavioral less stable, we hypothesized that the local dynamic stability, state. In certain conditions, variability may actually increase. would decrease in the dual-task condition. With respect to This contradictory relationship is noticeable when observ- movement variability, no particular effects were predicted ing the rich behavioral repertoire of elite sport players or because of contradictory or absence of earlier findings. Our expert musicians (Harbourne and Stergiou 2009; Glazier pilot study was conducted to increase our understanding on et al. 2003). how a cognitive load may contribute to an increased risk for An effective way to manipulate variability and stability in work-related overuse injuries, a topic to which we will return a cyclical motor task is adding a secondary cognitive task. in the “Discussion” section. This method is of particular interest in the context of the risks for work-related MSDs since cognitive demands are a relevant occupational factor which has been shown to affect Methods sustained repetitive movements (Srinivasan and Mathias- sen 2012; Bloemsaat et al. 2005). The underlying theory Participants of studies on dual-tasking is that resources are limited, and they have to be shared between a cognitive and a motor task, Thirteen right-handed healthy subjects (9 female, 4 male; consequently performance will suffer (Plummer and Eskes 25.46 ± 3.46 years) volunteered for this study. No participant 2015). In dual-task paradigms, local dynamical stability reported pain or history of injuries in neck, shoulder and might decrease, in terms of limited resources, since more arm regions. All participants gave informed written consent difficult tasks demand more resources and as a consequence 1 3 Experimental Brain Research (2018) 236:1611–1619 1613 and the study was approved by an institutionalized ethics the chair was then maintained in all trials. All participants review board. performed a warming-up trial for a minimum of 30 s or until they felt comfortable with the task. Procedure Equipment The protocol consisted of two trials of 5 min each, whose order was counterbalanced between participants. In one A custom-made script in Python 2.7 2010 (Python Soft- trial, a motor task was performed solo (M) and in the other ware Foundation, Beaverton, OR, USA) was implemented a motor task was performed in combination with a cogni- to present the stimuli and record the endpoint position on tive task (M + C). In the cognitive task, participants counted a 27′′ touch screen (1920 × 1080 resolution; ProLiteb Iiy- aloud backwards in multiples of three. In the motor task, ama, Iiyama Corporation, Tokyo, Japan). Four targets of participants performed a sustained repetitive task on a multi- 27 mm in diameter were presented, with a between-targets touch screen, tapping two pairs of visually presented targets distance of 125 mm in anterior direction and a distance of with both hands simultaneously and in-phase (Fig. 1). Par- 155 mm between the targets of the two hands. Xsens MVN ticipants could perform the task freely (without specified BIOMECH motion capture suit (Xsens technologies BV, rhythm or posture), with the only requirement of touching Enschede, The Netherlands) was used to record upper body the targets as fast and as accurate as possible. The motor task kinematics at 60 Hz. Eleven sensor units were placed on the shares common features with the bimanual Fitts’ task used head, sternum, pelvis, shoulders, upper arms, forearms and in previous studies (Longo and Meulenbroek 2018; Shea hands following the recommendations by Xsens. Anatomi- et al. 2012; Amazeen et al. 2005), however, no specific task cal measurements and calibrations were performed accord- variations as regards movement amplitude and target width, ing to the procedures provided by Xsens. Data acquisition were applied in the current study to enhance its monotony. was done via the accompanying software (MVN Studio 4.2, Before starting the measurements, participants were asked Xsens technologies BV, Enschede, Netherlands) which cal- to adjust the chair height and distance to the touch screen culates the kinematic data. Thirty joint angles: the 3D angle to find the most comfortable position. The configuration of configurations of wrist (2: left and right), elbow (2), shoul- der (2) and four column angles: C1–Head, T1–C7, T9–T8, L5–S1, were considered for further analysis. All joint angles were expressed in local coordinate systems following the guidelines of the International Society of Biomechanics (Wu et al. 2005). Data analysis Touch screen data (XY coordinates) of the fingertip posi - tions realized during the 5-min task were used to quantify the endpoint variability. We used the standard measure of variable error (VE), which was defined as the mean distance of all movement endpoints to the mean endpoint (Gordon et  al. 1994). For further calculations, we determined for each participant the mean of VE for all four targets (VE ). One subject was excluded from this analysis since data were missing. Xsens data were used to investigate other types of vari- ability and local dynamic stability, which were the primary interest in the present study. For this purpose, a PCA was applied using the 30 joint angles as 30-dimensional input vectors. Prior to data analysis, the first 5 s were excluded from the raw dataset, to avoid analyzing settling-in behav- ior. For each trial, every angle vector was normalized by subtracting the trial-mean. Then, a single input matrix was created with the normalized vectors as columns and the data of all subjects and both conditions (M, M + C) concat- enated vertically. Finally, a single PCA was calculated on Fig. 1 Experimental setup 1 3 1614 Experimental Brain Research (2018) 236:1611–1619 this combined input matrix to facilitate direct comparisons series from its trend was lower than half its total average between participants (Federolf 2016; Gløersen et al. 2017). (i.e., pause within the repetitive task), or if it exceeded two The first three principal components (PCs) were considered times its total average (e.g., unusual movement). The slope for the analysis of postural reconfigurations which reflect of the trends underlying the residual time periods were used the movement variability related to the volunteer’s postural to delineate transitions, non-stationary, and quasi-stationary changes ( see also Longo et al. 2018). The first principal phases, respectively. Specifically, transitions were defined if component was further examined for the analysis of cycle- the absolute value of the slope for a minimum of 100 sam- to-cycle variability and local dynamic stability in repetitive ples exceeded a threshold of 0.1 and non-stationary phases cycles. All calculations in the current study were imple- if the absolute value of the slope for a minimum of 300 sam- mented in Matlab R2015a (MathWorks Inc., Natic, MA, ples exceeded a threshold of 0.02. Thresholds and number USA). of samples used were specific for our setup and best identi- fied the four phases. Time periods that were not allocated to Postural reconfigurations any of the former phases were marked as quasi-stationary phases. If a criterion for any of the phases was met in one Changes in the postural configuration (Fig.  2) were deter- PC, then this period was delineated accordingly in all PC mined by first defining the trends of the first three PCs (black time series (Fig. 2). For further comparisons, the cumulated lines) through a low pass filter (Butterworth filter; cut-off duration per minute of each phase (D) was calculated. Thus, 0.1 Hz). Then, the trends were used to classify four phases: for each condition (M, M + C), four dependent variables were events—interruptions or unusual movements during the defined: D (events), D (transitions), D (non-stationary e t ns execution of the task; transitions—rapid changes from a phases), and D (quasi-stationary phases). qs postural configuration to another; non-stationary phases— gradual changes between postural configurations; quasi-sta- Cycle‑to‑cycle variability tionary phases—unchanging postural configurations. In par - ticular, events were defined by subtracting each trend from Thirty consecutive cycles were selected in the PC1 time its PC; an event was marked if the deviation of the PC time series in a quasi-stationary phase (Fig.  3a). The cycles M+C -20 -40 050 100 150 200 250 050100 150200 250 -20 -40 -20 -40 050100 150200 250 050 100 150200 250 time [s] Fig. 2 Representative dataset of a 5  min trial of the motor (M) and stationary phases (cyan), non-stationary phases (green), and transi- the motor + cognitive (M + C) trial of one arbitrarily selected volun- tions (red). The black line represents the low pass-filtered underlying teer: the first three PCs are shown. The tapping movement between trend two pairs of targets is printed as a colored line, respectively, quasi- 1 3 PC3 PC1 PC2 Experimental Brain Research (2018) 236:1611–1619 1615 selected corresponded to the first quasi-stationary phase exponential rate of separation of neighboring trajectories of of at least 30 cycles, i.e., longest consecutive cycles that the attractor. LyE was calculated by first constructing a state could be detected among all participants and both condi- space representation of the time series (Fig. 3b). The time tions in the relevant time periods. A cycle was defined delay (τ) was determined using the average mutual infor- as a back and forward movement, starting from the tar- mation (AMI; Fraser and Swinney 1986 )and the embed- gets closer to the body. The starting points of the cycles ding dimension (m) using a false nearest neighbor algorithm corresponded to the local maxima of PC1. Spatial (SD ) (Kantz 1994) Therefore, m = 2 and τ = 9 were selected. and temporal (SD ) variability were calculated on the 30 Finally, LyE values were calculated for the time series using selected cycles. SD was calculated by first interpolating Wolf’s algorithm (Wolf et al. 1985; Buzzi et al. 2003.) each cycle such that it was represented by 100 samples (i.e., expressed in percent). For each sample, the stand- Statistical analysis ard deviation between cycles was determined. Finally, the mean of the standard deviations over the whole cycle was To determine changes in the postural configuration, the calculated. SD was assessed as the standard deviation of cumulated duration per minute of each phase (D , D , D , T e t ns the movement duration between cycles. The mean of the D ) was compared between the two conditions (M, M + C). qs movement duration (T ) between the 30-selected cycles As the data were not normally distributed, a Wilcoxon was also assessed as a control measure. signed-rank test was used. Variables D , D , D are inde- e t ns pendent and were analyzed applying a Šidák correction for Local dynamic stability multiple comparisons, thus reducing the α-level for sta- tistical significance to α = 0.0174. For completeness, also The largest Lyapunov exponent (LyE) was calculated for variable D , which directly depends on the other variables qs the same 30 cycles selected in the PC1-time series. LyE is (D  = 60 s − [D  + D  + D ]) was analyzed; also applying qs e t ns a measure of local dynamic stability, which quantifies the the corrected α-threshold of α = 0.0174. For the analysis of M+C (a) -20 -40 050100 150200 250 050 100 150 200 250 time [s] time [s] (b) LyE = LyE =3.42 LyE =1.97 Fig. 3 a Representation of PC1 of the motor (M) and the motor + cog- ysis of cycle-to-cycle variability. b State space representation of 30 nitive (M + C) trial of one arbitrary selected subject. The enlargement cycles of the same representative subject for the analysis of the larg- shows 30 cycles selected in the quasi-stationary phases for the anal- est Lyapunov exponent 1 3 PC1 1616 Experimental Brain Research (2018) 236:1611–1619 cycle-to-cycle variability, local dynamic stability and end- to PC1 and were largely affected by postural reconfigurations point variability the data were normally distributed, there- of the subjects. fore, a paired-samples t test was used to compare SD SD , C, T T , LyE, and VE for both conditions. Here, the α-level for Motor task versus motor + cognitive task m m statistical significance was set to α = 0.05. Statistical analy- ses were performed using SPSS Version 22 (IBM, Chicago, With respect to endpoint variability, a statistical trend IL, USA). was observed in VE which decreased in the M + C [9.34 (± 3.22) mm] compared to the M [11.01 (± 3.78) mm] trials [t(11) = 1.83, P = 0.095, d = 0.48]. Postural reconfigurations Results (Fig. 4a) revealed a significant main effect in D which was qs higher in the M + C than in the M trial (Z = 2.43, P = 0.015, Results of the principal component analysis r = 0.67). A statistical trend was found in D (Z = 2.02, P = 0.043, r = 0.56), indicating more frequent changes in the Principal components 1–3 represented 44.9, 16.9, and 9.9% M than in the M + C trial. No significant differences between of the overall variance in the kinematic data, respectively. conditions were found in D (Z = 1.73, P = 0.084, r = 0.48) ns Figure 2 shows an example of the first three PC score time and in D (Z = 0.67, P = 0.5, r = 0.19). series of the M and M + C trial for one selected subject. The Cycle-to-cycle variability (Fig.  4b) revealed a signifi- first principal component (PC1) represented the movement cant main effect between conditions in SD [t(12) = 2.39, component containing the largest variance and, in the cur- P = 0.036, d = 0.75] indicating higher temporal variability rent study, PC1 was dominated by the cyclic movement pat- in the M + C than in the M trial. However, no significant tern of the task. PC2 and PC3 represent variance orthogonal differences between conditions were observed in SD Changes in the postural configuration (a) * 2 25 25 20 20 1.5 15 15 10 10 0.5 5 5 0 0 0 0 MM+C MM+C MM+C MM+C Cycle-to-cycle variability Local dynamic stability (b) (c) 0.08 0.06 0.04 3 0.02 M M+C M M+C MM+C Fig. 4 a Box plots of cumulative duration per minute of events (D ; (SD ) and temporal variability (SD ) for M and M + C; c box plot of e C T magenta), transitions (D ; red), non-stationary phases (D ; green) the largest Lyapunov exponent (LyE) for M and M + C. Significant t ns and quasi-stationary phases (D ; cyan)for the motor (M) and the between condition effects are indicated by an asterisk qs motor + cognitive (M + C) trial; b box plots of spatial variability 1 3 D [s] SD D [s] SD [s] D [s] ns LyE D [s] qs Experimental Brain Research (2018) 236:1611–1619 1617 [t(12) = 0.45, P = 0.664, d = 0.16] and in T [t(12) = 1.24, due to a concurrent cognitive task, as we observed in the P = 0.239, d = 0.33]. Local dynamical stability (Fig.  4c) current study, local stability may be reduced. A dynamical decreased in the M + C as compared to the M trial as movement system can try to attenuate local fluctuations and reflected by an increase of  LyE [t (12) = 2.36, P = 0.036, maintain a stable coordination pattern by adopting another d = 0.6]. functional solution or coordination mode that suits the dual- task constraints better. The result of this process may be that the system is constrained at the joint level thus reducing the Discussion incidence of postural reconfigurations. A novelty of the present study is the application of PCA The current pilot study explored the effects of a concurrent for the assessment of different types of variability and local- cognitive task on different types of movement variability, dynamic stability. Using this approach, we moved away from i.e., endpoint variability, postural reconfiguration, cycle- quantyfing the variability of isolated joints by a limited to-cycle variability, and on the local dynamic stability, i.e., number of pre-selected kinematic variables, and instead, largest Lyapunov exponent, in a sustained repetitive upper- moved towards metrics such as postural reconfigurations of extremity motor task. In agreement with our hypothesis and the whole upper body which allowed us to capture complex the view that tasks that demand more resources are less sta- multijoint coordination and thus provide a fuller account of ble, the local dynamic stability decreased under dual-task multijoint cyclical movements while coping with a cogni- conditions (Fig. 4c). However, the effects of the secondary tive load. Further, we attempted to better understand what cognitive load on different types of movement variability distinct parameters measuring variability and stability reflect revealed contrasting results as compared to earlier studies in sustained upper-extremity motion. Our results show that that used different variability measures (Beauchet et  al. LyE and temporal variability reflect unwanted fluctuations 2005; Hollman et al. 2007). Temporal variability increased in performance due to reduced control with an increase in (Fig. 4b), suggesting that the cognitive task caused interfer- task difficulty.The incidence of postural reconfigurations, ence due to the competition for attentional resources neces- however, reflects a potential beneficial variability due to the sary for the motor task. The increase of temporal variability dynamics of the human movement system. Another benefit with an additional cognitive load is in line with dual-task of distinguishing between different types of variability by interference effects reported earlier (Dubost et  al. 2006; means of PCA is that the LyE can then be calculated on Beauchet et al. 2005). Spatial variability was not affected quasi-stationary phases. Stationarity of the underling time by the counting task and the endpoint variability marginally series is a prerequisite for this calculation, but in human decreased. Simultaneously, the incidence of postural recon- movement studies this stationarity is often difficult to define. gu fi rations signica fi ntly decreased in the dual-task condition However, one limitation of this approach was that the num- (Fig.4a), indicating that participants adopted fixed postures ber of consecutive cycles needed for the calculation of LyE for longer periods of time. Since motor variability has been was limited by the occurrence of postural reconfigurations, purported as beneficial for avoiding overuse injuries and nonetheless the length selected is considered adequate for pain (Bartlett et al. 2007; Srinivasan and Mathiassen 2012), the analysis (cf. Wolf et al. 1985). Further, due to the novelty the decrease in postural readjustments due to dual tasking of the current approach and the low sample size, our findings may constitute a risk factor for MSDs. need to be taken with caution. The postural readjustment results may be interpreted In conclusion, the current findings suggest that under from the viewpoint of dynamical systems theory as follows. cognitive demands, the temporal variability and dynamic Generally, in dual-task paradigms, the challenge for a motor instability of cyclical arm movements increase. Simultane- system performing movements and a cognitive task is to ously, at the postural level, cognitive loads led to a decreased adapt to the secondary task demands without reducing the incidence of postural reconfigurations. Particularly, for the quality of movement performance. The main purpose of a prevention of MSDs, this reduced postural variability should dynamical system then is to reach or maintain global sta- be carefully monitored since postural reconfigurations may bility. Goal-directed actions are supported by reducing the play a role in the prevention of overuse injuries. number of biomechanical degrees of freedom of the motor Acknowledgements This study was n fi ancially supported by the Euro - system through the formation of functional synergies afford- pean Union FP7 Marie Curie IDP Grant (FP7-PEOPLE-2013-ITN ing preferred and stable coordination patterns. However, a ‘HealthPAC’, Grant 604063-IDP). We wish to acknowledge the sup- stable system does permit flexible and adaptive motor behav - port of Henk Luinge and Matteo Giuberti (Xsens technologies BV, Enschede, The Netherlands) for providing the sensors and helping with ior, encouraging free exploration of coordination changes the preparation of the experiment. to be able to acquire different stable motor solutions over time, a mechanism known to enhance motor learning (Gla- zier et  al. 2003). Now, if the motor system is perturbed 1 3 1618 Experimental Brain Research (2018) 236:1611–1619 Dhawale AK, Smith MA, Olveczky BP (2017) The role of variability Compliance with ethical standards in motor learning. Annu Rev Neurosci 40:479–498. https ://doi. org/10.1146/annur ev-neuro -07211 6 Conflict of interest The authors have no conflicts of interest. Dingwell JB, Marin LC (2006) Kinematic variability and local dynamic stability of upper body motions when walking at differ - Ethical standards The study was approved by an institutionalized ethics ent speeds. J Biomech 39(3):444–452. https ://doi.org/10.1016/j. review board in accordance with the ethical standards laid down in the jbiom ech.2004.12.014 1964 Declaration of Helsinki. Dingwell JB, Cusumano JP, Cavanagh PR, Sternad D (2001) Local dynamic stability versus kinematic variability of continuous Informed consent All participants gave informed written consent prior overground and treadmill walking. J Biomech Eng 123(1):27–32. to their inclusion in the study. http://www.ncbi.nlm.nih.gov/pubme d/11277 298 Dubost V, Kressig RW, Gonthier R, Herrmann FR, Aminian K, Najafi B, Beauchet O (2006) Relationships between dual-task Open Access This article is distributed under the terms of the Crea- related changes in stride velocity and stride time variability in tive Commons Attribution 4.0 International License (http://creat iveco healthy older adults. Hum Mov Sci 25(3):372–382. https ://doi. mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- org/10.1016/j.humov .2006.03.004 tion, and reproduction in any medium, provided you give appropriate Federolf PA (2016) A novel approach to study human posture control: credit to the original author(s) and the source, provide a link to the “Principal movements” obtained from a principal component Creative Commons license, and indicate if changes were made. analysis of kinematic marker data. J Biomech 49(3):364–370. https ://doi.org/10.1016/j.jbiom ech.2015.12.030 Federolf P, Tecante K, Nigg B (2012) A holistic approach to study the temporal variability in gait. J Biomech 45(7):1127–1132. https :// References doi.org/10.1016/j.jbiom ech.2012.02.008 Fraser AM, Swinney HL (1986) Independent coordinates for strange Amazeen EL, Ringenbach SD, Amazeen PG (2005) The effects of attractors from mutual information. Phys Rev A 33(2):1134–1140. attention and handedness on coordination dynamics in a biman- https ://doi.org/10.1103/PhysR evA.33.1134 ual Fitts’ law task. Exp Brain Res 164(4):484–499. https ://doi. Fuller JR, Fung J, Côté JN (2011) Time-dependent adaptations to org/10.1007/s0022 1-005-2269-y posture and movement characteristics during the development of Bartlett R, Wheat J, Robins M (2007) Is movement variability impor- repetitive reaching induced fatigue. Exp Brain Res 211(1):133– tant for sports biomechanists? Sports Biomech 6(2):224–243. 143. https ://doi.org/10.1007/s0022 1-011-2661-8 https ://doi.org/10.1080/14763 14070 13229 94 Glazier PS, Davids K, Bartlett RM (2003) Dynamical systems theory: a Beauchet O, Dubost V, Herrmann FR, Kressig RW (2005) relevant framework for performance-oriented sports biomechanics Stride-to-stride variability while backward counting among research. Sportscience. http://shura .shu.ac.uk/13427 /. Accessed healthy young adults. J NeuroEng Rehabil. https ://doi. 2 Feb 2018 org/10.1186/1743-0003-2-26 Gløersen Ø, Myklebust H, Hallén J, Federolf P (2017) Technique anal- Bernstein N (1967) The co-ordination and regulation of movements. ysis in elite athletes using principal component analysis. J Sports Pergamon Press, Oxford Sci. https ://doi.org/10.1080/02640 414.2017.12988 26 Beurskens R, Bock O (2013) Does the walking task matter? Influ- Gordon J, Ghilardi MF, Cooper SE, Ghez C (1994) Accuracy of pla- ence of different walking conditions on dual-task performances nar reaching movements I. Independence of direction and extent in young and older persons. Hum Mov Sci 32:1456–1466. https variability. Exp Brain Res 99(1):112–130. https://doi.or g/10.1007/ ://doi.org/10.1016/j.humov .2013.07.013 BF002 41416 Bloemsaat JG, Meulenbroek RGJ, Van Galen GP (2005) Differential Hak L, Houdijk H, Beek PJ, Van Dieën JH (2013) Steps to take to effects of mental load on proximal and distal arm muscle activ - enhance gait stability: the effect of stride frequency, stride length, ity. Exp Brain Res 167(4):622–634. https://doi.or g/10.1007/s0022 and walking speed on local dynamic stability and margins of 1-005-0066-2 stability. PloS One 8(12):e82842. https ://doi.org/10.1371/journ Buzzi UH, Stergiou N, Kurz MJ, Hageman PA, Heidel J (2003) Non- al.pone.00828 42 linear dynamics indicates aging affects variability during gait. Hamill J, Palmer C, Van Emmerik REA (2012) Coordinative vari- Clin Biomech 18(5):435–443. https ://doi.or g/10.1016/S0268 ability and overuse injury. SMARTT 4(1):45. https ://doi. -0033(03)00029 -9 org/10.1186/1758-2555-4-45 Cignetti F, Schena F, Rouard A (2009) Effects of fatigue on inter-cycle Harbourne RT, Stergiou N (2009) Movement variability and the use variability in cross-country skiing. J Biomech 42(10):1452–1459. of nonlinear tools: principles to guide physical therapist practice. https ://doi.org/10.1016/j.jbiom ech.2009.04.012 Phys Ther 89(3):267–282. https ://doi.org/10.2522/ptj.20080 130 Côté JN, Raymond D, Mathieu PA, Feldman AG, Levin MF (2005) Hollman JH, Kovash FM, Kubik JJ, Linbo RA (2007) Age-related Differences in multi-joint kinematic patterns of repetitive ham- differences in spatiotemporal markers of gait stability dur- mering in healthy, fatigued and shoulder-injured individuals. Clin ing dual task walking. Gait Posture 26(1):113–119. https ://doi. Biomech 20(6):581–590. https ://doi.or g/10.1016/j.clinb iomec org/10.1016/j.gaitp ost.2006.08.005 h.2005.02.012 Kantz H (1994) A robust method to estimate the maximal Lyapunov Daffertshofer A, Lamoth CJC, Meijer OG, Beek PJ (2004) PCA in exponent of a time series. Phys Lett A 185(1):77–87. https ://doi. studying coordination and variability: a tutorial. Clin Bio- org/10.1016/0375-9601(94)90991 -1 mech 19(4):415–428. ht tp s : //d oi .o r g/ 10 .1 01 6/ j.c l inb i om ec Kelso JAS (1995) Dynamic patterns: the self-organization of brain and h.2004.01.005 behavior. The MIT Press, Cambridge Danion F, Bongers RM, Bootsma RJ (2014) The trade-off between Longo A, Meulenbroek R (2018) Precision-dependent changes in spatial and temporal variabilities in reciprocal upper-limb aiming motor variability during sustained bimanual reaching. Motor movements of different durations. PloS One 9(5):e97447. https: // Control 22:28–44. https ://doi.org/10.1123/mc.2016-0013 doi.org/10.1371/journ al.pone.00974 47 Longo A, Meulenbroek R, Haid T, Federolf P (2018) Postural reconfig- uration and cycle-to-cycle variability in patients with work-related 1 3 Experimental Brain Research (2018) 236:1611–1619 1619 musculoskeletal disorders compared to healthy controls and in Stergiou N, Decker LM (2011) Human movement variability, nonlin- relation to pain emerging during a repetitive movement task. ear dynamics, and pathology: is there a connection? Hum Mov Clin Biomech 54:103–110. https ://doi.org/10.1016/j.clinb iomec Sci 30(5):869–888. https://doi.or g/10.1016/j.humov.2011.06.002 h.2018.03.004 Van Emmerik REA, Ducharme SW, Amado AC, Hamill J (2016) Com- Madeleine P, Mathiassen SE, Arendt-Nielsen L (2008) Changes in paring dynamical systems concepts and techniques for biome- the degree of motor variability associated with experimental chanical analysis. J Sport Health Sci. https ://doi.org/10.1016/j. and chronic neck-shoulder pain during a standardised repeti- jshs.2016.01.013 tive arm movement. Exp Brain Res 185(4):689–698. https ://doi. Wolf A, Swift JB, Swinney HL, Vastano JA (1985) Determining Lya- org/10.1007/s0022 1-007-1199-2 punov exponents from a time series. Phys D Nonlinear Phenom Magnani RM, Lehnen GC, Rodrigues FB, de Sá e Souza GS, de 16(3):285–317. https ://doi.org/10.1016/0167-2789(85)90011 -9 Oliveira Andrade A, Vieira MF (2017) Local dynamic sta- Wolpert DM, Ghahramani Z, Flanagan JR (2001) Perspectives and bility and gait variability during attentional tasks in young problems in motor learning. Trends Cognit Sci 5(11):487–494. adults. Gait Posture 55:105–108. https ://doi.org/10.1016/j.gaitp https ://doi.org/10.1016/S1364 -6613(00)01773 -3 ost.2017.04.019 Woollacott M, Shumway-Cook A (2002) Attention and the control Plummer P, Eskes G (2015) Measuring treatment effects on dual-task of posture and gait: a review of an emerging area of research. performance: a framework for research and clinical practice. Front Gait Posture 16(1):1–14. https ://doi.or g/10.1016/S0966 Hum Neurosci 9:225. https://doi.or g/10.3389/fnhum.2015.00225 -6362(01)00156 -4 Segal AD, Orendurff MS, Czerniecki JM, Shofer JB, Klute GK Wu G, van der Helm FCT, Veeger HEJD, Makhsous M, Van Roy P, (2008) Local dynamic stability in turning and straight-line gait. Anglin C, International Society of Biomechanics, et al (2005) J Biomech 41(7):1486–1493. https ://doi.org/10.1016/J.JBIOM ISB recommendation on definitions of joint coordinate systems ECH.2008.02.012 of various joints for the reporting of human joint motion—part II: Shea CH, Boyle J, Kovacs AJ (2012) Bimanual Fitts’ tasks: Kelso, Sou- shoulder, elbow, wrist and hand. J Biomech 38(5):981–992. https thard, and Goodman, 1979 revisited. Exp Brain Res 216(1):113–://doi.org/10.1016/J.JBIOM ECH.2004.05.042 121. https ://doi.org/10.1007/s0022 1-011-2915-5 Srinivasan D, Mathiassen SE (2012) Motor variability in occupational health and performance. Clin Biomech. https: //doi.org/10.1016/j. clinb iomec h.2012.08.007 1 3

Journal

Experimental Brain ResearchSpringer Journals

Published: Mar 27, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off