Gait & Posture 39 (2014) 1146–1148
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Towards clinical application: Repetitive sensor position re-calibration for improved reliability of gait parameters Daniel Hamacher a,1,*, Dennis Hamacher a,1, William R. Taylor b,1, Navrag B. Singh b,1, Lutz Schega a,1 a b
Department of Sport Science, Otto von Guericke University Magdeburg, Brandenburger Str. 9, Magdeburg 39104, Germany Institute for Biomechanics, ETH Zu¨rich, Wolfgang-Pauli-Str. 10, Zurich 8093, Switzerland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 29 July 2013 Received in revised form 14 January 2014 Accepted 26 January 2014
While camera-based motion tracking systems are considered to be the gold standard for kinematic analysis, these systems are not practical in clinical practice. However, the collection of gait parameters using inertial sensors is feasible in clinical settings and less expensive, but suffers from drift error that excludes accurate analyses. The goal of this study was to apply a combination of repetitive sensor position re-calibration techniques in order to improve the intra-day and inter-day reliability of gait parameters using inertial sensors. Kinematic data of nineteen healthy elderly individuals were captured twice within the first day and once on a second day after one week using inertial sensors fixed on the subject’s forefoot during gait. Parameters of walking speed, minimum foot clearance (MFC), minimum toe clearance (MTC), stride length, stance time and swing time, as well as their corresponding measures of variability were calculated. Intra-day and inter-day differences were rated using intra-class correlation coefficients (ICC(3,1)), as well as the bias and limits of agreement. The results indicate excellent reliability for all intra-day and inter-day mean parameters (ICC: MFC 0.83–stride length 0.99). While good to excellent reliability was observed during intra-day parameters of variability (ICC: walking speed 0.71–MTC 0.98), corresponding inter-day reliability ranged from poor to excellent (ICC: walking speed 0.32–MTC 0.95). In conclusion, the system is suitable for reliable measurement of mean temporo-spatial parameters and the variability of MFC and MTC. However, the system’s accuracy needs to be improved before remaining parameters of variability can reliably be collected. ß 2014 Elsevier B.V. All rights reserved.
Keywords: Inertial sensors Gait analysis Temporo-spatial parameters Gait variability Reliability
1. Introduction Clinical gait analysis provides new options in the accurate assessment of subtle changes in gait characteristics associated with motor pathologies [1]. The evaluation of gait variability is known to provide key information for differentiating subjects with neuromuscular deficits from their healthy counterparts and therefore aid in the early clinical diagnosis of disease or the prediction of future fallers [2]. Camera-based motion capture systems are deemed the gold standard for conducting kinematic gait analyses. However, the
* Corresponding author. Tel.: +49 3916756600; fax: +49 3916716754. E-mail address: [email protected] (D. Hamacher). 1 All authors were fully involved in the study and preparation of the manuscript. Each of the authors has read and concurs with the content in the final manuscript. The material within has not been and will not be submitted for publication elsewhere except as an abstract. http://dx.doi.org/10.1016/j.gaitpost.2014.01.020 0966-6362/ß 2014 Elsevier B.V. All rights reserved.
expense, time and expertise required for capture and analysis of the data, and the need for a well-equipped laboratory prevent the wider applicability of such systems in clinical practice [3]. The recent introduction of inertial sensors for the low cost and rapid collection of multiple repetitions of gait parameters [3] now offers options for assessing normal overground walking in clinical settings. While the concept of using inertial sensors for gait analyses has already been introduced [4], their efficacy has not yet been sufficiently demonstrated with respect to clinical application. One disadvantage of inertial sensors is that they do not have a fixed reference point, which introduces errors due to drift. Hence, effective correction needs to be introduced before accurate sensor position can be determined. Moreover, temporo-spatial parameters are calculated using double integration of the acceleration data, possibly explaining why the reliability of these parameters and their variability are not yet sufficiently resolved. By applying approaches to reduce drift through repetitive re-calibration at each stride, the aim of the current study was to address the issue of
D. Hamacher et al. / Gait & Posture 39 (2014) 1146–1148
reliable sensor position and assess the reliability of the resulting temporo-spatial parameters of gait and their variability using inertial sensors in an elderly cohort. 2. Methods 2.1. Subjects Gait data of 19 healthy elderly subjects (5 male, 19 female, age: 71 4 years) were captured twice within the first day and once after seven days. All subjects provided written informed consent to participate in this study. 2.2. Testing procedure A wireless inertial motion tracker (MTw sensors, Xsens Technologies B.V., Netherlands) was fixed to each of the subjects’ forefeet. Kinematic data was then captured while the subjects walked continuously five times up and down a level hallway (distance 25 m) at their preferred walking speed, and steadily turn around two pylons in a radius of about 1 m.
1147
the global coordinate system (GCS), ~ as;g was transformed into the GCS using Q. Afterwards, gravitational acceleration was removed by subtracting 9.81 m/s2 from the vertical acceleration to derive ~ a. The velocity of the foot-attached sensor was assumed to be temporarily equal to zero during FF [9], which served as a boundary condition for the step-wise integration within two subsequent FF phases of foot acceleration (~ a) in order to access ~ n. To minimise the drift in velocity, ~ n at FF was re-set to zero at each new stride, which was thus used as a repetitive start condition and for error backpropagation with the aid of a linear drift model [7]. To calculate the foot (sensor) position ~ p, the vertical foot position (pvertical) during FF was re-set to zero at each new stride. Afterwards, the minimum foot clearance (MFC) was defined as the local minimum vertical sensor position during the swing phase of gait. In addition, foot orientations and sensor heights at TO were used to calculate the toe position relative to the sensor [10]. The median of the relative toe position was used to derive the toe trajectory. The minimum toe clearance (MTC) was analogously derived from the MFC together with the orientation of the sensor. As a result, the following gait parameters were determined: mean of walking speed, stride length, stride time, stance time, swing time, cadence, MFC, and MTC, as well as each parameters’ coefficient of variation (CoV).
2.3. Analysis 2.4. Statistics The first and last 2.5 m of each 25 m section were removed to avoid transients. Since the reliability of variability measures are known to depend upon the number of strides analysed [5], we incorporated 100 strides of each foot per participant. An in-house algorithm was then developed to calculate a range of gait parameters, including a specific approach to allow repeated re-calibration of the sensors’ position and avoid progressive accumulation of position drift errors: Using the MTw sensors, the 3D orientations (determined as quaternions, Q), 3D calibrated accelerations (~ as;g , in the local sensor coordinate system; LSCS), as ~ , also in the LSCS), well as the 3D calibrated angular velocities (v were all collected at a frequency of 75 Hz. Foot positions were deduced from foot acceleration data [4], and the gait events of heel ~ in the sagittal contact (HC) and toe-off (TO), local minima of v plane, were extracted [6,7]. Furthermore, the flat foot (FF) phase ~ in the sagittal plane between was defined as the lowest absolute v HC and TO [8]. To extract movement related foot accelerations in
For estimating intra- and inter-day reliability, intra-class correlation coefficients, ICC(3,1), were calculated (IBM SPSS Statistics 20) [11], where values between 0.0–0.40 were considered poor, 0.40–0.59 fair, 0.60–0.74 good, and 0.75–1.00 to be excellent [12]. Furthermore, we calculated the Bias and the Limits of Agreement (LoA, [13]) as measures of agreement between trials. 3. Results Intra-day ICCs from 0.983 (MTC, left foot) to 0.996 (stride length, right foot) were determined, thus displaying excellent reliability (Table 1). Additionally, very low biases (e.g. stride length: 0.001 m, MFC: 0.000 m) and small LoA (e.g. stride length left/right: 0.028/0.026 m, MFC left/right: 0.002/0.001 m) were observed. All variability measures apart from the CoV of walking speed (ICC: 0.711) showed excellent reliability, exhibiting a range
Table 1 Intra-day and inter-day reliability of gait parameters and their variability of the left/right foot (Coefficient of variation: CoV, Limit of Agreement: LoA, minimum foot clearance: MFC, minimum toe clearance: MTC, excellent: ex). Mean
Intra-day reliability BIAS (left/right)
Stride length (m) MFC (m) MTC (m) Walking speed (m/s) Cadence (Steps/s) Stride time (s) Stance time (s) Swing time (s) CoV
0.001/ 0.001 0.000/0.000 0.001/0.001 0.002/0.002 0.002/0.002 0.001/ 0.002 0.001/ 0.001 0.000/ 0.001
LoA (left/right)
ICC (left/right)
(left/right)
0.028/0.026 0.002/0.001 0.002/0.001 0.048/0.047 0.016/0.016 0.015/0.015 0.009/0.010 0.007/0.007
0.991/0.996 0.990/0.993 0.983/0.987 0.988/0.989 0.992/0.993 0.993/0.944 0.994/0.994 0.989/0.993
ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex
Intra-day reliability BIAS (left/right)
Stride length (%) MFC (%) MTC (%) Walking speed (%) Cadence (%) Stride time (%) Stance time (%) Swing time (%)
Inter-day reliability
0.002/0.144 0.002/ 0.305 0.053/ 0.812 0.460/0.205 0.000/0.072 0.000/0.069 0.115/0.079 0.000/0.067
BIAS (left/right) 0.023/0.033 0.001/0.000 0.001/0.000 0.031/0.048 0.006/0.011 0.005/ 0.011 0.000/ 0.008 0.005/ 0.003
LoA (left/right)
(left/right)
(left/right)
0.076/0.057 0.006/0.005 0.006/0.006 0.144/0.113 0.059/0.052 0.063/0.053 0.049/0.032 0.025/0.024
0.943/0.945 0.834/0.873 0.865/0.836 0.880/0.877 0.904/0.923 0.890/0.913 0.841/0.861 0.893/0.929
ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex
Inter-day reliability LoA (left/right)
ICC (left/right)
(left/right)
0.027/0.549 0.006/3.398 0.141/4.958 3.331/0.916 0.008/0.489 0.001/0.491 0.392/0.399 0.002/0.566
0.799/0.791 0.974/0.933 0.975/0.907 0.711/0.809 0.854/0.803 0.859/0.809 0.926/0.897 0.928/0.862
ex/ex ex/ex ex/ex good/ex ex/ex ex/ex ex/ex ex/ex
BIAS (left/right) 0.138/ 0.102 0.374/ 0.141 1.344/0.763 0.319/ 0.163 0.208/ 0.116 0.211/ 0.123 0.365/ 0.281 0.195/ 0.127
LoA (left/right) 1.042/0.831 6.968/6.674 11.020/9.369 1.254/1.808 0.521/0.882 0.521/0.915 0.931/1.146 0.841/0.795
ICC (left/right) 0.465/0.567 0.795/0.781 0.812/0.951 0.493/0.320 0.554/0.476 0.542/0.450 0.337/0.361 0.350/0.604
(left/right) fair/fair ex/ex ex/ex fair/poor fair/fair fair/fair poor/poor poor/good
[(Fig._1)TD$IG]
1148
D. Hamacher et al. / Gait & Posture 39 (2014) 1146–1148
Fig. 1. Scheme of the calculation procedure of gait parameters.
from ICC = 0.791 (stride length, right) to ICC = 0.975 (MTC) with a bias of 0.002%/0.144% (left/right) and LoA of 0.027% 0.549% (left/right) for stride length. The results of the inter-day analyses yielded excellent ICCs ranging from 0.834 to 0.945 for mean parameters of gait. However, the reliability of inter-day measures of variability was only excellent for MFC (ICC left/right: 0.795/0.781) and MTC (ICC left/right: 0.812/0.951). 4. Discussion This study aimed to clarify how the analysis of gait using inertial sensors can be improved through the application of recalibration techniques for use in clinical settings. While the intraday reliability indicated almost exclusively excellent reliability of the variability measures, inter-day reliability was excellent only for CoV-MFC and CoV-MTC, possibly due to the variation of sensor reattachment positions. To our knowledge, only one test-retest study has examined the intra-day reliability of gait parameters using inertial sensors fixed to the lower limbs [14], which reported worse ICC values (e.g. CoV walking speed: 0.50, CoV stride time: 0.1) than here. In our study, mean parameters of gait showed good-excellent reliability throughout, possibly due to the large number of strides, which is in agreement with the literature [15]. The lower reliability of stride-to-stride variability, however, was probably a result of the relatively low measurement frequency of 75 Hz, where integration of the data introduces large errors. Here, higher frequencies would improve the integration accuracy of the acceleration signal. This particularly affects variability measures, since SD captures deviations from the mean and therefore combines motor variability but also measurement system error. In conclusion, the proposed calculation procedure (Fig. 1), which includes a novel combination of algorithms and allows repeated re-calibration of sensor position data, is suitable for reliably measuring mean parameters of gait, as well as the variability of MFC and MTC.
Conflict of interests There is no conflict of interests. References [1] Whittle MW. Clinical gait analysis: a review. Hum Movement Sci 1996; 15(3):369–87. [2] Hamacher D, Singh NB, Van Diee¨n JH, Heller MO, Taylor WR. Kinematic measures for assessing gait stability in elderly individuals: a systematic review. J R Soc Interf 2011;8(65):1682–98. [3] Reininga IHF, Stevens M, Wagenmakers R, Bulstra SK, Groothoff JW, Zijlstra W. Subjects with hip osteoarthritis show distinctive patterns of trunk movements during gait-a body-fixed-sensor based analysis. J Neuroeng Rehabil 2012;9(1). [4] Mariani B, Hoskovec C, Rochat S, Bu¨la C, Penders J, Aminian K. 3D gait assessment in young and elderly subjects using foot-worn inertial sensors. J Biomech 2010;43(15):2999–3006. [5] Hollman JH, Childs KB, McNeil ML, Mueller AC, Quilter CM, Youdas JW. Number of strides required for reliable measurements of pace, rhythm and variability parameters of gait during normal and dual task walking in older individuals. Gait Posture 2010;32(1):23–8. [6] Mariani B, Rouhani H, Crevoisier X, Aminian K. Quantitative estimation of footflat and stance phase of gait using foot-worn inertial sensors. Gait Posture 2013;37(2):229–34. [7] Sabatini AM, Martelloni C, Scapellato S, Cavallo F. Assessment of walking features from foot inertial sensing. IEEE T Bio-Med Eng 2005;52(3):486–94. [8] Zhu S, Anderson H, Wang Y, A. Real-time on-chip algorithm for IMU-based gait measurement. Lect Notes Comput Sci 2012;93–104. [9] Peruzzi A, Della Croce U, Cereatti A. Estimation of stride length in level walking using an inertial measurement unit attached to the foot: a validation of the zero velocity assumption during stance. J Biomech 2011;44(10):1991–4. [10] Mariani B, Rochat S, Bu¨la CJ, Aminian K. Heel and toe clearance estimation for gait analysis using wireless inertial sensors. IEEE T Bio-Med Eng 2012;59(12 Part 2):3162–8. [11] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86(2):420–8. [12] Cicchetti DV. Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychol Assessment 1994;6(4):284–90. [13] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1(8476):307–10. [14] Najafi B, Helbostad JL, Moe-Nilssen R, Zijlstra W, Aminian K. Does walking strategy in older people change as a function of walking distance? Gait Posture 2009;29(2):261–6. [15] Ko¨nig N, Singh NB, von Beckerath J, Janke L, Taylor WR. Is gait variability reliable? An assessment of spatio-temporal parameters of gait variability during continuous overground walking. Gait Posture 2014;39(1):615–7.
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Towards clinical application: Repetitive sensor position re-calibration for improved reliability of gait parameters Daniel Hamacher a,1,*, Dennis Hamacher a,1, William R. Taylor b,1, Navrag B. Singh b,1, Lutz Schega a,1 a b
Department of Sport Science, Otto von Guericke University Magdeburg, Brandenburger Str. 9, Magdeburg 39104, Germany Institute for Biomechanics, ETH Zu¨rich, Wolfgang-Pauli-Str. 10, Zurich 8093, Switzerland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 29 July 2013 Received in revised form 14 January 2014 Accepted 26 January 2014
While camera-based motion tracking systems are considered to be the gold standard for kinematic analysis, these systems are not practical in clinical practice. However, the collection of gait parameters using inertial sensors is feasible in clinical settings and less expensive, but suffers from drift error that excludes accurate analyses. The goal of this study was to apply a combination of repetitive sensor position re-calibration techniques in order to improve the intra-day and inter-day reliability of gait parameters using inertial sensors. Kinematic data of nineteen healthy elderly individuals were captured twice within the first day and once on a second day after one week using inertial sensors fixed on the subject’s forefoot during gait. Parameters of walking speed, minimum foot clearance (MFC), minimum toe clearance (MTC), stride length, stance time and swing time, as well as their corresponding measures of variability were calculated. Intra-day and inter-day differences were rated using intra-class correlation coefficients (ICC(3,1)), as well as the bias and limits of agreement. The results indicate excellent reliability for all intra-day and inter-day mean parameters (ICC: MFC 0.83–stride length 0.99). While good to excellent reliability was observed during intra-day parameters of variability (ICC: walking speed 0.71–MTC 0.98), corresponding inter-day reliability ranged from poor to excellent (ICC: walking speed 0.32–MTC 0.95). In conclusion, the system is suitable for reliable measurement of mean temporo-spatial parameters and the variability of MFC and MTC. However, the system’s accuracy needs to be improved before remaining parameters of variability can reliably be collected. ß 2014 Elsevier B.V. All rights reserved.
Keywords: Inertial sensors Gait analysis Temporo-spatial parameters Gait variability Reliability
1. Introduction Clinical gait analysis provides new options in the accurate assessment of subtle changes in gait characteristics associated with motor pathologies [1]. The evaluation of gait variability is known to provide key information for differentiating subjects with neuromuscular deficits from their healthy counterparts and therefore aid in the early clinical diagnosis of disease or the prediction of future fallers [2]. Camera-based motion capture systems are deemed the gold standard for conducting kinematic gait analyses. However, the
* Corresponding author. Tel.: +49 3916756600; fax: +49 3916716754. E-mail address: [email protected] (D. Hamacher). 1 All authors were fully involved in the study and preparation of the manuscript. Each of the authors has read and concurs with the content in the final manuscript. The material within has not been and will not be submitted for publication elsewhere except as an abstract. http://dx.doi.org/10.1016/j.gaitpost.2014.01.020 0966-6362/ß 2014 Elsevier B.V. All rights reserved.
expense, time and expertise required for capture and analysis of the data, and the need for a well-equipped laboratory prevent the wider applicability of such systems in clinical practice [3]. The recent introduction of inertial sensors for the low cost and rapid collection of multiple repetitions of gait parameters [3] now offers options for assessing normal overground walking in clinical settings. While the concept of using inertial sensors for gait analyses has already been introduced [4], their efficacy has not yet been sufficiently demonstrated with respect to clinical application. One disadvantage of inertial sensors is that they do not have a fixed reference point, which introduces errors due to drift. Hence, effective correction needs to be introduced before accurate sensor position can be determined. Moreover, temporo-spatial parameters are calculated using double integration of the acceleration data, possibly explaining why the reliability of these parameters and their variability are not yet sufficiently resolved. By applying approaches to reduce drift through repetitive re-calibration at each stride, the aim of the current study was to address the issue of
D. Hamacher et al. / Gait & Posture 39 (2014) 1146–1148
reliable sensor position and assess the reliability of the resulting temporo-spatial parameters of gait and their variability using inertial sensors in an elderly cohort. 2. Methods 2.1. Subjects Gait data of 19 healthy elderly subjects (5 male, 19 female, age: 71 4 years) were captured twice within the first day and once after seven days. All subjects provided written informed consent to participate in this study. 2.2. Testing procedure A wireless inertial motion tracker (MTw sensors, Xsens Technologies B.V., Netherlands) was fixed to each of the subjects’ forefeet. Kinematic data was then captured while the subjects walked continuously five times up and down a level hallway (distance 25 m) at their preferred walking speed, and steadily turn around two pylons in a radius of about 1 m.
1147
the global coordinate system (GCS), ~ as;g was transformed into the GCS using Q. Afterwards, gravitational acceleration was removed by subtracting 9.81 m/s2 from the vertical acceleration to derive ~ a. The velocity of the foot-attached sensor was assumed to be temporarily equal to zero during FF [9], which served as a boundary condition for the step-wise integration within two subsequent FF phases of foot acceleration (~ a) in order to access ~ n. To minimise the drift in velocity, ~ n at FF was re-set to zero at each new stride, which was thus used as a repetitive start condition and for error backpropagation with the aid of a linear drift model [7]. To calculate the foot (sensor) position ~ p, the vertical foot position (pvertical) during FF was re-set to zero at each new stride. Afterwards, the minimum foot clearance (MFC) was defined as the local minimum vertical sensor position during the swing phase of gait. In addition, foot orientations and sensor heights at TO were used to calculate the toe position relative to the sensor [10]. The median of the relative toe position was used to derive the toe trajectory. The minimum toe clearance (MTC) was analogously derived from the MFC together with the orientation of the sensor. As a result, the following gait parameters were determined: mean of walking speed, stride length, stride time, stance time, swing time, cadence, MFC, and MTC, as well as each parameters’ coefficient of variation (CoV).
2.3. Analysis 2.4. Statistics The first and last 2.5 m of each 25 m section were removed to avoid transients. Since the reliability of variability measures are known to depend upon the number of strides analysed [5], we incorporated 100 strides of each foot per participant. An in-house algorithm was then developed to calculate a range of gait parameters, including a specific approach to allow repeated re-calibration of the sensors’ position and avoid progressive accumulation of position drift errors: Using the MTw sensors, the 3D orientations (determined as quaternions, Q), 3D calibrated accelerations (~ as;g , in the local sensor coordinate system; LSCS), as ~ , also in the LSCS), well as the 3D calibrated angular velocities (v were all collected at a frequency of 75 Hz. Foot positions were deduced from foot acceleration data [4], and the gait events of heel ~ in the sagittal contact (HC) and toe-off (TO), local minima of v plane, were extracted [6,7]. Furthermore, the flat foot (FF) phase ~ in the sagittal plane between was defined as the lowest absolute v HC and TO [8]. To extract movement related foot accelerations in
For estimating intra- and inter-day reliability, intra-class correlation coefficients, ICC(3,1), were calculated (IBM SPSS Statistics 20) [11], where values between 0.0–0.40 were considered poor, 0.40–0.59 fair, 0.60–0.74 good, and 0.75–1.00 to be excellent [12]. Furthermore, we calculated the Bias and the Limits of Agreement (LoA, [13]) as measures of agreement between trials. 3. Results Intra-day ICCs from 0.983 (MTC, left foot) to 0.996 (stride length, right foot) were determined, thus displaying excellent reliability (Table 1). Additionally, very low biases (e.g. stride length: 0.001 m, MFC: 0.000 m) and small LoA (e.g. stride length left/right: 0.028/0.026 m, MFC left/right: 0.002/0.001 m) were observed. All variability measures apart from the CoV of walking speed (ICC: 0.711) showed excellent reliability, exhibiting a range
Table 1 Intra-day and inter-day reliability of gait parameters and their variability of the left/right foot (Coefficient of variation: CoV, Limit of Agreement: LoA, minimum foot clearance: MFC, minimum toe clearance: MTC, excellent: ex). Mean
Intra-day reliability BIAS (left/right)
Stride length (m) MFC (m) MTC (m) Walking speed (m/s) Cadence (Steps/s) Stride time (s) Stance time (s) Swing time (s) CoV
0.001/ 0.001 0.000/0.000 0.001/0.001 0.002/0.002 0.002/0.002 0.001/ 0.002 0.001/ 0.001 0.000/ 0.001
LoA (left/right)
ICC (left/right)
(left/right)
0.028/0.026 0.002/0.001 0.002/0.001 0.048/0.047 0.016/0.016 0.015/0.015 0.009/0.010 0.007/0.007
0.991/0.996 0.990/0.993 0.983/0.987 0.988/0.989 0.992/0.993 0.993/0.944 0.994/0.994 0.989/0.993
ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex
Intra-day reliability BIAS (left/right)
Stride length (%) MFC (%) MTC (%) Walking speed (%) Cadence (%) Stride time (%) Stance time (%) Swing time (%)
Inter-day reliability
0.002/0.144 0.002/ 0.305 0.053/ 0.812 0.460/0.205 0.000/0.072 0.000/0.069 0.115/0.079 0.000/0.067
BIAS (left/right) 0.023/0.033 0.001/0.000 0.001/0.000 0.031/0.048 0.006/0.011 0.005/ 0.011 0.000/ 0.008 0.005/ 0.003
LoA (left/right)
(left/right)
(left/right)
0.076/0.057 0.006/0.005 0.006/0.006 0.144/0.113 0.059/0.052 0.063/0.053 0.049/0.032 0.025/0.024
0.943/0.945 0.834/0.873 0.865/0.836 0.880/0.877 0.904/0.923 0.890/0.913 0.841/0.861 0.893/0.929
ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex ex/ex
Inter-day reliability LoA (left/right)
ICC (left/right)
(left/right)
0.027/0.549 0.006/3.398 0.141/4.958 3.331/0.916 0.008/0.489 0.001/0.491 0.392/0.399 0.002/0.566
0.799/0.791 0.974/0.933 0.975/0.907 0.711/0.809 0.854/0.803 0.859/0.809 0.926/0.897 0.928/0.862
ex/ex ex/ex ex/ex good/ex ex/ex ex/ex ex/ex ex/ex
BIAS (left/right) 0.138/ 0.102 0.374/ 0.141 1.344/0.763 0.319/ 0.163 0.208/ 0.116 0.211/ 0.123 0.365/ 0.281 0.195/ 0.127
LoA (left/right) 1.042/0.831 6.968/6.674 11.020/9.369 1.254/1.808 0.521/0.882 0.521/0.915 0.931/1.146 0.841/0.795
ICC (left/right) 0.465/0.567 0.795/0.781 0.812/0.951 0.493/0.320 0.554/0.476 0.542/0.450 0.337/0.361 0.350/0.604
(left/right) fair/fair ex/ex ex/ex fair/poor fair/fair fair/fair poor/poor poor/good
[(Fig._1)TD$IG]
1148
D. Hamacher et al. / Gait & Posture 39 (2014) 1146–1148
Fig. 1. Scheme of the calculation procedure of gait parameters.
from ICC = 0.791 (stride length, right) to ICC = 0.975 (MTC) with a bias of 0.002%/0.144% (left/right) and LoA of 0.027% 0.549% (left/right) for stride length. The results of the inter-day analyses yielded excellent ICCs ranging from 0.834 to 0.945 for mean parameters of gait. However, the reliability of inter-day measures of variability was only excellent for MFC (ICC left/right: 0.795/0.781) and MTC (ICC left/right: 0.812/0.951). 4. Discussion This study aimed to clarify how the analysis of gait using inertial sensors can be improved through the application of recalibration techniques for use in clinical settings. While the intraday reliability indicated almost exclusively excellent reliability of the variability measures, inter-day reliability was excellent only for CoV-MFC and CoV-MTC, possibly due to the variation of sensor reattachment positions. To our knowledge, only one test-retest study has examined the intra-day reliability of gait parameters using inertial sensors fixed to the lower limbs [14], which reported worse ICC values (e.g. CoV walking speed: 0.50, CoV stride time: 0.1) than here. In our study, mean parameters of gait showed good-excellent reliability throughout, possibly due to the large number of strides, which is in agreement with the literature [15]. The lower reliability of stride-to-stride variability, however, was probably a result of the relatively low measurement frequency of 75 Hz, where integration of the data introduces large errors. Here, higher frequencies would improve the integration accuracy of the acceleration signal. This particularly affects variability measures, since SD captures deviations from the mean and therefore combines motor variability but also measurement system error. In conclusion, the proposed calculation procedure (Fig. 1), which includes a novel combination of algorithms and allows repeated re-calibration of sensor position data, is suitable for reliably measuring mean parameters of gait, as well as the variability of MFC and MTC.
Conflict of interests There is no conflict of interests. References [1] Whittle MW. Clinical gait analysis: a review. Hum Movement Sci 1996; 15(3):369–87. [2] Hamacher D, Singh NB, Van Diee¨n JH, Heller MO, Taylor WR. Kinematic measures for assessing gait stability in elderly individuals: a systematic review. J R Soc Interf 2011;8(65):1682–98. [3] Reininga IHF, Stevens M, Wagenmakers R, Bulstra SK, Groothoff JW, Zijlstra W. Subjects with hip osteoarthritis show distinctive patterns of trunk movements during gait-a body-fixed-sensor based analysis. J Neuroeng Rehabil 2012;9(1). [4] Mariani B, Hoskovec C, Rochat S, Bu¨la C, Penders J, Aminian K. 3D gait assessment in young and elderly subjects using foot-worn inertial sensors. J Biomech 2010;43(15):2999–3006. [5] Hollman JH, Childs KB, McNeil ML, Mueller AC, Quilter CM, Youdas JW. Number of strides required for reliable measurements of pace, rhythm and variability parameters of gait during normal and dual task walking in older individuals. Gait Posture 2010;32(1):23–8. [6] Mariani B, Rouhani H, Crevoisier X, Aminian K. Quantitative estimation of footflat and stance phase of gait using foot-worn inertial sensors. Gait Posture 2013;37(2):229–34. [7] Sabatini AM, Martelloni C, Scapellato S, Cavallo F. Assessment of walking features from foot inertial sensing. IEEE T Bio-Med Eng 2005;52(3):486–94. [8] Zhu S, Anderson H, Wang Y, A. Real-time on-chip algorithm for IMU-based gait measurement. Lect Notes Comput Sci 2012;93–104. [9] Peruzzi A, Della Croce U, Cereatti A. Estimation of stride length in level walking using an inertial measurement unit attached to the foot: a validation of the zero velocity assumption during stance. J Biomech 2011;44(10):1991–4. [10] Mariani B, Rochat S, Bu¨la CJ, Aminian K. Heel and toe clearance estimation for gait analysis using wireless inertial sensors. IEEE T Bio-Med Eng 2012;59(12 Part 2):3162–8. [11] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86(2):420–8. [12] Cicchetti DV. Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychol Assessment 1994;6(4):284–90. [13] Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1(8476):307–10. [14] Najafi B, Helbostad JL, Moe-Nilssen R, Zijlstra W, Aminian K. Does walking strategy in older people change as a function of walking distance? Gait Posture 2009;29(2):261–6. [15] Ko¨nig N, Singh NB, von Beckerath J, Janke L, Taylor WR. Is gait variability reliable? An assessment of spatio-temporal parameters of gait variability during continuous overground walking. Gait Posture 2014;39(1):615–7.
