- Email: [email protected]
Effects of age and walking speed on long-range autocorrelations and on fluctuation magnitude of stride duration
S44
Abstracts / Gait & Posture 36 (2012) S1–S101
References
References
[1] McGinley, et al. Gait Posture 2009;29:360–9. [2] Bland, Altman. Lancet 1986;1:307–10.
[1] Engsberg, et al. Gait & Posture 2009;29:169–71. [2] Gorton, et al. Gait & Posture 2009;29:398–402. [3] Ferrari, et al. Gait & Posture 2008;28(2):207–16.
doi:10.1016/j.gaitpost.2011.10.253 doi:10.1016/j.gaitpost.2011.10.254 O73 O74 Inter-laboratory consistency of gait analysis measurements M.G. Benedetti, A. Merlo, M. Boschi, A. Leardini ∗
Effects of age and walking speed on long-range autocorrelations and on fluctuation magnitude of stride duration
Movement Analysis Laboratory, Istituto Ortopedico Rizzoli, Bologna, Italy
B. Bollens 1,∗ , C. Detrembleur 1 , F. Crevecoeur 2 , T. Lejeune 1
Introduction: Sharing data among motion analysis laboratories is a top priority in gait analysis research [1]. Repeatability of the typical gait analysis measurements has been rarely assessed over laboratories [2]. Our purpose is to explore the consistency of routine gait analysis on a single healthy subject once it is performed in different clinical laboratories. Patients/materials and methods: A single healthy subject, 26year-old, male, 177 cm height, 70 kg weight, was examined within a 1-month period in seven gait analysis laboratories. These had their own motion analysis systems, and established protocols: 4 ‘conventional’, 2 Total3DGait, 1 CAST. A full kinematics and kinetics assessment was performed at each site by a local examiner. Each lab provided six gait trials data, along with other characteristics (pathway length, force-plate positions, details on hardware, software releases, kinematics and analog data sampling rate, etc.). The consistency of all the measurements was assessed by the coefficient of variation (CV) and the maximum spread, i.e. the worst-case difference. These were used for the anthropometric measurements, spatio-temporal parameters, local peaks, braking and propulsive areas from the GRF, and all joint rotation values in up-right posture (‘offset’). The similarity between two curves was assessed by the coefficient of determination r2 , along with the scaling factor (SF) and the bias term (B), obtained by a robust linear regression, thus separating the effect of the vertical shifts from that of shape variation. For any given gait analysis curve, all possible 21 two-lab comparisons were performed and the obtained coefficients averaged. This procedure was applied to left and right lower limb joint rotations, moments and powers in the sagittal plane. Results: Differences as large as 2–3 cm were found in the anthropometric measurements at the pelvis. CV of the spatio-temporal parameters was in general lower than 6%. GRF were similar, with differences in the order of a few percentage of the body weight. The similarity amongst joint kinematic curves was high, with, on average, r2 > 0.90 in both the sagittal and the frontal planes and r2 > 0.60 in the transverse plane (knee excluded) with the worst performance at the hip (r2 = 0.30). Pattern similarity was excellent for joint moments at the ankle (r2 = 0.90) and good at knee (r2 = 0.70) and hip (r2 = 0.66) joints. For the latter, comparisons between labs with the same protocols resulted in better similarities (r2 > 0.80). Pattern similarity was very good also for joint power at the ankle (r2 = 0.80) but not satisfactory at the knee and the hip joints (r2 = 0.30 and r2 = 0.40 respectively). Discussion and conclusion: Large consistency was found in joint kinematic and kinetic curves, despite of the large spectrum of differences in the techniques utilized in the laboratories. Different protocols can result in similar curves, except for those measurements where knee axes and hip centre are implied [3]. Relevant variability was observed to be accounted for differences in markers positioning on the thigh, anthropometric measures, and event detection, all dependent on the examiner. Disclosure: No significant relationships.
1
Institute of Neuroscience, Université Catholique de Louvain, Brussels, Belgium 2 Queen’s University, Kingston, Canada Introduction: Stride duration variability is often considered as a marker of gait balance and can be investigated on two different ways [1,2]. Classical mathematical methods (standard deviations, coefficients of variation) assess fluctuation magnitude, whereas fluctuation dynamics (long-range autocorrelations) is evaluated through complex mathematical methods. Both variables were reported as different in children and in older adults in comparison with young adults. However, the effects of age on stride duration variability are on one hand mostly assessed at spontaneous speed, which is always different between groups. On the other hand, gait speed is described as a possible confounder in the assessment of stride duration variability, but its precise influence remains controversial. The first objective of this study was to assess the effect of gait speed on stride duration variability of young healthy subjects. The second aim was to investigate the influence of age on the same parameters. Patients/materials and methods: Stride duration was measured with a high sampling rate thanks to footswitches [3]. We first compared data obtained from 6 young healthy subjects (19–24 years old) walking during several minutes (512 consecutive gait cycles) at 6 different speeds (20–40–70–100–130–160% of spontaneous speed) on a treadmill (One Way RM ANOVA). Second, we compared the results of 18 subjects from 3 different age groups (∼5–25–75 years old) walking at three equivalent speeds (slow–medium–fast) (One Way ANOVA). Fluctuation magnitude was assessed by coefficients of variation (CV), while fluctuation dynamics was evaluated thanks to gold-standard methods (rescaled range analysis, power spectral analysis, relation d) [4], to reach a high level of evidence. Results: Values of CV were inversely related to speed and age: they were significantly greater at slowest speeds in young subjects (p < 0.001), and were also significantly larger for children in comparison with young and old adults, whatever the speed of walking (p = 0.011 at slow speed; p = 0.002 at medium speed; and p < 0.001 at fast speed). Conversely, long-range autocorrelations were found in all subjects at all speeds. Moreover, gait speed and age of subjects did not influence their characteristics, that remained similar in the different testing conditions. Discussion and conclusion: These results confirm with a high level of evidence that long-range autocorrelations are quite robust and intrinsic to the locomotor system. Fluctuation magnitude and dynamics are differently influenced by speed and age, which indicates that they probably have different neurophysiological origins. The precise clinical implications of those results need further investigation, in particular regarding fall risk, but CV and long-range autocorrelations could be complementary tools in the assessment of gait balance. Disclosure: No significant relationships.
Abstracts / Gait & Posture 36 (2012) S1–S101
S45
References [1] Hausdorff JM, et al. Is walking a random walk? Evidence for long-range correlations in stride interval of human gait. J Appl Physiol 1995. [2] Jordan K, et al. Walking speed influences on gait cycle variability. Gait Posture 2007. [3] Bollens B, et al. Does human gait exhibit comparable and reproducible longrange autocorrelations on level ground and on treadmill? Gait Posture 2010. [4] Crevecoeur F, et al. Towards a “gold-standard” approach to address the presence of long-range auto-correlation in physiological time series. J Neurosci Methods 2010.
doi:10.1016/j.gaitpost.2011.10.255 O75 Are stability issues a potential cause for adaptations in split belt walking? S.M. Bruijn ∗ , J. Duysens, S. Swinnen Department of Biomedical Kinesiology, Motor Control Laboratory, Research Center for Movement Control and Neuroplasticity, K.U. Leuven, Leuven, Belgium Introduction: Although the split belt paradigm has been used extensively to assess the capability of subjects to adapt the gait pattern [1], reasons as to why this adaptation occurs remain obscure. For stride length, which quickly adapts, this is clear; if no adaptation would take place, subjects would not be able to keep walking on the treadmill. For step length however, this is not the case; in principle, subjects are capable of a pattern that is not adapted; they show such a pattern in the immediate phase after the transition from tied to split belts. Still, they adapt their gait pattern, and, store this adapted gait pattern, so that after-effects are visible. This leads one to believe that the adapted gait pattern is in some way “more optimal” than the unadapted gait pattern. Potential ways in which the adapted gait pattern may be “more optimal” could be symmetry, energy consumption and stability. In the current study, we tested the latter of these, to see if gait adaptations during split belt walking are aimed at stabilizing the gait pattern. We hypothesized that (1) unadapted split belt walking (as seen in the initial phases of the adaptation period), would be less stable than normal walking and (2) that stability would improve as the gait pattern got more adapted to the split belt. Patients/materials and methods: 8 healthy subjects participated in a split belt adaptation protocol. In short, they walked on a treadmill for 5 min with belts tied (running at 1.0 m/s), then 10 min with belts split (1.0 and 0.5 m/s). Kinematics of a pelvis cluster marker were recorded using a optoelectronic measurement system at 100 samples/s. Time series of the 3d velocity and angular displacement of the pelvis marker were cut into episodes of 15 strides, and of each episode, stability was estimated using the maximum Lyapunov exponent (s ) [2]. Differences in stability between normal walking (designated as the average over all episodes during the tied condition), initial adaptation (average value over the first 5 episodes of split belt walking) and late adaptation (average value over episodes 26–30 of split belt walking) were tested using a repeated measures ANOVA. Results: Results are shown in Fig. 1. In line with our hypotheses, split belt walking initially decreased stability (i.e. led to higher values of s ), but stability improved as the gait pattern adapted to split belt walking, so that during the late adaptation period, stability was not significantly different from normal walking. Discussion and conclusion: The present study suggests that stability issues may be responsible for the adaptations in split belt walking. Other factors may also improve because of these adaptations. One likely candidate is energy consumption [3]. We plan on further exploring the potential reasons why humans adapt their gait pattern when walking on a split belt.
Fig. 1.
Disclosure: No significant relationships. References [1] Reisman DS, et al. J Neurophysiol 2005;94(4):2403–15. [2] Bruijn SM, et al. J Neurosci Methods 2009;178(2):327–33. [3] Zarrugh MY, et al. Eur J Appl Physiol Occup Physiol 1974;33(4):293–306.
doi:10.1016/j.gaitpost.2011.10.256 O76 Asymmetry vs. deviation plot: A new gait analysis data reduction tool N.G. Darras 1 , D. Pasparakis 2,∗ , M. Tziomaki 1 , C. Nestoridis 1 , M. Pentarakis 1 1
Hellenic Society of Disabled Children, ELEPAP Gait Lab, Athens, Greece 2nd Orthopaedic Department, “Aglaia Kyriakou”, Children Hospital, Athens, Greece 2
Introduction: Clinical practice in gait analysis, accepts a range of deviation and asymmetry that is considered as normal. This range is defined by one normal SD around the normal mean values. In order to summarize the gait condition of a patient, various attempts have been made, using either the deviation from normal by calculating a deviation index, or for evaluating asymmetry, asymmetry/symmetry indexes. Most often asymmetry is not being considered and the evaluation focuses on deviations from normal. During the application of the indexes in the clinical practice, for assessing the condition of a subject’s gait, it occurred that the evaluation of deviations from normal and the inherent asymmetry
Abstracts / Gait & Posture 36 (2012) S1–S101
References
References
[1] McGinley, et al. Gait Posture 2009;29:360–9. [2] Bland, Altman. Lancet 1986;1:307–10.
[1] Engsberg, et al. Gait & Posture 2009;29:169–71. [2] Gorton, et al. Gait & Posture 2009;29:398–402. [3] Ferrari, et al. Gait & Posture 2008;28(2):207–16.
doi:10.1016/j.gaitpost.2011.10.253 doi:10.1016/j.gaitpost.2011.10.254 O73 O74 Inter-laboratory consistency of gait analysis measurements M.G. Benedetti, A. Merlo, M. Boschi, A. Leardini ∗
Effects of age and walking speed on long-range autocorrelations and on fluctuation magnitude of stride duration
Movement Analysis Laboratory, Istituto Ortopedico Rizzoli, Bologna, Italy
B. Bollens 1,∗ , C. Detrembleur 1 , F. Crevecoeur 2 , T. Lejeune 1
Introduction: Sharing data among motion analysis laboratories is a top priority in gait analysis research [1]. Repeatability of the typical gait analysis measurements has been rarely assessed over laboratories [2]. Our purpose is to explore the consistency of routine gait analysis on a single healthy subject once it is performed in different clinical laboratories. Patients/materials and methods: A single healthy subject, 26year-old, male, 177 cm height, 70 kg weight, was examined within a 1-month period in seven gait analysis laboratories. These had their own motion analysis systems, and established protocols: 4 ‘conventional’, 2 Total3DGait, 1 CAST. A full kinematics and kinetics assessment was performed at each site by a local examiner. Each lab provided six gait trials data, along with other characteristics (pathway length, force-plate positions, details on hardware, software releases, kinematics and analog data sampling rate, etc.). The consistency of all the measurements was assessed by the coefficient of variation (CV) and the maximum spread, i.e. the worst-case difference. These were used for the anthropometric measurements, spatio-temporal parameters, local peaks, braking and propulsive areas from the GRF, and all joint rotation values in up-right posture (‘offset’). The similarity between two curves was assessed by the coefficient of determination r2 , along with the scaling factor (SF) and the bias term (B), obtained by a robust linear regression, thus separating the effect of the vertical shifts from that of shape variation. For any given gait analysis curve, all possible 21 two-lab comparisons were performed and the obtained coefficients averaged. This procedure was applied to left and right lower limb joint rotations, moments and powers in the sagittal plane. Results: Differences as large as 2–3 cm were found in the anthropometric measurements at the pelvis. CV of the spatio-temporal parameters was in general lower than 6%. GRF were similar, with differences in the order of a few percentage of the body weight. The similarity amongst joint kinematic curves was high, with, on average, r2 > 0.90 in both the sagittal and the frontal planes and r2 > 0.60 in the transverse plane (knee excluded) with the worst performance at the hip (r2 = 0.30). Pattern similarity was excellent for joint moments at the ankle (r2 = 0.90) and good at knee (r2 = 0.70) and hip (r2 = 0.66) joints. For the latter, comparisons between labs with the same protocols resulted in better similarities (r2 > 0.80). Pattern similarity was very good also for joint power at the ankle (r2 = 0.80) but not satisfactory at the knee and the hip joints (r2 = 0.30 and r2 = 0.40 respectively). Discussion and conclusion: Large consistency was found in joint kinematic and kinetic curves, despite of the large spectrum of differences in the techniques utilized in the laboratories. Different protocols can result in similar curves, except for those measurements where knee axes and hip centre are implied [3]. Relevant variability was observed to be accounted for differences in markers positioning on the thigh, anthropometric measures, and event detection, all dependent on the examiner. Disclosure: No significant relationships.
1
Institute of Neuroscience, Université Catholique de Louvain, Brussels, Belgium 2 Queen’s University, Kingston, Canada Introduction: Stride duration variability is often considered as a marker of gait balance and can be investigated on two different ways [1,2]. Classical mathematical methods (standard deviations, coefficients of variation) assess fluctuation magnitude, whereas fluctuation dynamics (long-range autocorrelations) is evaluated through complex mathematical methods. Both variables were reported as different in children and in older adults in comparison with young adults. However, the effects of age on stride duration variability are on one hand mostly assessed at spontaneous speed, which is always different between groups. On the other hand, gait speed is described as a possible confounder in the assessment of stride duration variability, but its precise influence remains controversial. The first objective of this study was to assess the effect of gait speed on stride duration variability of young healthy subjects. The second aim was to investigate the influence of age on the same parameters. Patients/materials and methods: Stride duration was measured with a high sampling rate thanks to footswitches [3]. We first compared data obtained from 6 young healthy subjects (19–24 years old) walking during several minutes (512 consecutive gait cycles) at 6 different speeds (20–40–70–100–130–160% of spontaneous speed) on a treadmill (One Way RM ANOVA). Second, we compared the results of 18 subjects from 3 different age groups (∼5–25–75 years old) walking at three equivalent speeds (slow–medium–fast) (One Way ANOVA). Fluctuation magnitude was assessed by coefficients of variation (CV), while fluctuation dynamics was evaluated thanks to gold-standard methods (rescaled range analysis, power spectral analysis, relation d) [4], to reach a high level of evidence. Results: Values of CV were inversely related to speed and age: they were significantly greater at slowest speeds in young subjects (p < 0.001), and were also significantly larger for children in comparison with young and old adults, whatever the speed of walking (p = 0.011 at slow speed; p = 0.002 at medium speed; and p < 0.001 at fast speed). Conversely, long-range autocorrelations were found in all subjects at all speeds. Moreover, gait speed and age of subjects did not influence their characteristics, that remained similar in the different testing conditions. Discussion and conclusion: These results confirm with a high level of evidence that long-range autocorrelations are quite robust and intrinsic to the locomotor system. Fluctuation magnitude and dynamics are differently influenced by speed and age, which indicates that they probably have different neurophysiological origins. The precise clinical implications of those results need further investigation, in particular regarding fall risk, but CV and long-range autocorrelations could be complementary tools in the assessment of gait balance. Disclosure: No significant relationships.
Abstracts / Gait & Posture 36 (2012) S1–S101
S45
References [1] Hausdorff JM, et al. Is walking a random walk? Evidence for long-range correlations in stride interval of human gait. J Appl Physiol 1995. [2] Jordan K, et al. Walking speed influences on gait cycle variability. Gait Posture 2007. [3] Bollens B, et al. Does human gait exhibit comparable and reproducible longrange autocorrelations on level ground and on treadmill? Gait Posture 2010. [4] Crevecoeur F, et al. Towards a “gold-standard” approach to address the presence of long-range auto-correlation in physiological time series. J Neurosci Methods 2010.
doi:10.1016/j.gaitpost.2011.10.255 O75 Are stability issues a potential cause for adaptations in split belt walking? S.M. Bruijn ∗ , J. Duysens, S. Swinnen Department of Biomedical Kinesiology, Motor Control Laboratory, Research Center for Movement Control and Neuroplasticity, K.U. Leuven, Leuven, Belgium Introduction: Although the split belt paradigm has been used extensively to assess the capability of subjects to adapt the gait pattern [1], reasons as to why this adaptation occurs remain obscure. For stride length, which quickly adapts, this is clear; if no adaptation would take place, subjects would not be able to keep walking on the treadmill. For step length however, this is not the case; in principle, subjects are capable of a pattern that is not adapted; they show such a pattern in the immediate phase after the transition from tied to split belts. Still, they adapt their gait pattern, and, store this adapted gait pattern, so that after-effects are visible. This leads one to believe that the adapted gait pattern is in some way “more optimal” than the unadapted gait pattern. Potential ways in which the adapted gait pattern may be “more optimal” could be symmetry, energy consumption and stability. In the current study, we tested the latter of these, to see if gait adaptations during split belt walking are aimed at stabilizing the gait pattern. We hypothesized that (1) unadapted split belt walking (as seen in the initial phases of the adaptation period), would be less stable than normal walking and (2) that stability would improve as the gait pattern got more adapted to the split belt. Patients/materials and methods: 8 healthy subjects participated in a split belt adaptation protocol. In short, they walked on a treadmill for 5 min with belts tied (running at 1.0 m/s), then 10 min with belts split (1.0 and 0.5 m/s). Kinematics of a pelvis cluster marker were recorded using a optoelectronic measurement system at 100 samples/s. Time series of the 3d velocity and angular displacement of the pelvis marker were cut into episodes of 15 strides, and of each episode, stability was estimated using the maximum Lyapunov exponent (s ) [2]. Differences in stability between normal walking (designated as the average over all episodes during the tied condition), initial adaptation (average value over the first 5 episodes of split belt walking) and late adaptation (average value over episodes 26–30 of split belt walking) were tested using a repeated measures ANOVA. Results: Results are shown in Fig. 1. In line with our hypotheses, split belt walking initially decreased stability (i.e. led to higher values of s ), but stability improved as the gait pattern adapted to split belt walking, so that during the late adaptation period, stability was not significantly different from normal walking. Discussion and conclusion: The present study suggests that stability issues may be responsible for the adaptations in split belt walking. Other factors may also improve because of these adaptations. One likely candidate is energy consumption [3]. We plan on further exploring the potential reasons why humans adapt their gait pattern when walking on a split belt.
Fig. 1.
Disclosure: No significant relationships. References [1] Reisman DS, et al. J Neurophysiol 2005;94(4):2403–15. [2] Bruijn SM, et al. J Neurosci Methods 2009;178(2):327–33. [3] Zarrugh MY, et al. Eur J Appl Physiol Occup Physiol 1974;33(4):293–306.
doi:10.1016/j.gaitpost.2011.10.256 O76 Asymmetry vs. deviation plot: A new gait analysis data reduction tool N.G. Darras 1 , D. Pasparakis 2,∗ , M. Tziomaki 1 , C. Nestoridis 1 , M. Pentarakis 1 1
Hellenic Society of Disabled Children, ELEPAP Gait Lab, Athens, Greece 2nd Orthopaedic Department, “Aglaia Kyriakou”, Children Hospital, Athens, Greece 2
Introduction: Clinical practice in gait analysis, accepts a range of deviation and asymmetry that is considered as normal. This range is defined by one normal SD around the normal mean values. In order to summarize the gait condition of a patient, various attempts have been made, using either the deviation from normal by calculating a deviation index, or for evaluating asymmetry, asymmetry/symmetry indexes. Most often asymmetry is not being considered and the evaluation focuses on deviations from normal. During the application of the indexes in the clinical practice, for assessing the condition of a subject’s gait, it occurred that the evaluation of deviations from normal and the inherent asymmetry