Korean Physical Therapy Science
[ Article ]
The Journal of Korean Academy of Physical Therapy Science - Vol. 27, No. 3, pp.25-34
ISSN: 2733-6441 (Print) 2733-645X (Online)
Print publication date 31 Dec 2020
Received 06 Aug 2020 Revised 13 Sep 2020 Accepted 21 Sep 2020
DOI: https://doi.org/10.26862/jkpts.2020.12.27.3.25

The test-retest reliability of gait kinematic data measured using a portable gait analysis system in healthy adults

Jung-Ae An, M.Sc., D.P.T., P.T.1, 2 ; Kyung-Seok Byun, Ph.D.3 ; Byounghee Lee, Ph.D., P.T.4
1Dept. of Physical Therapy, Seoul Now Hospital
2Dept. of International, Korean organization of physical therapy in BodyDesign
3Vector Bio.
4Dept. of Physical Therapy, Sahmyook University

Correspondence to: 이병희 주소: 서울특별시 노원구 화랑로 815, 삼육대학교 제3과학관 112호, 전화: 02-3399-1634, E-mail: 3679@syu.ac.kr

© 2019 by the Korean Physical Therapy Science

Abstract

Background

Gait analysis is an important measurement for health professionals to assess gait patterns related to functional limitations due to neurological or orthopedic conditions. The purpose of this study was to investigate the reliability of the newly developed portable gait analysis system (PGAS).

Design

Cross-sectional design. Test-retest study.

Methods

The PGAS study was based on a wearable sensor, and measurement of gait kinematic parameters, such as gait velocity, cadence, step length and stride length, and joint angle (hip, knee, and ankle) in stance and swing phases. The results were compared with a motion capture system (MCS). Twenty healthy individuals were applied to the MCS and PGAS simultaneously during gait performance.

Results

The test-retest reliability of the PGAS showed good repeatability in gait parameters with mean intra-class correlation coefficients (ICCs) ranging from 0.840 to 0.992, and joint angles in stance and swing phase from 0.907 to 0.988. The acceptable test-retest ICC was observed for the gait parameters (0.809 to 0.961), and joint angles (0.800 to 0.977).

Conclusion

The results of this study indicated that the developed PGAS showed good grades of repeatability for gait kinematic data along with acceptable ICCs compared with the results from the MCS. The gait kinematic parameters in healthy subjects can be used as standard values for adopting this PGAS.

Keywords:

Gait analysis, Portable gait analysis system, Test-retest, Reliability

Ⅰ. Introduction

Over the past centuries, gait analysis has evolved by integrating measurements, descriptions, and evaluations to analyze human locomotion characteristics and provide information on gait kinematic and kinetic for researchers and clinicians (Ghoussayni et al, 2004). Gait analysis provides a baseline for planning, evaluation, and management for patient treatment in the medical field, also is used in elderly healthy to provide a valuable information (Dobkin et al, 2004; Yavuzer et al, 2008).

Gait analysis systems have been developed with its various needs, among which, due to the accuracy of the equipment, motion analysis systems have become the gold standard today but, these motion analysis systems required expensive high-speed cameras, specialized internal motion analysis laboratories, and post-processing periods to obtain accurate data, therefore, it was necessary to develop alternative gait analysis equipment using wearable sensors such as the accelerometer, gyroscope, magneto-resistive sensors, flexible goniometer, electromagnetic tracking system (ETS), sensing fabric, force sensor, and sensors for electromyography to overcome the limitations of the laboratory environment and high price (Tao et al, 2012; Wagner and Ganz, 2012; Park et al., 2019).

In recent years, various gait analysis devices based on wearable sensors have been proposed, and meaningful progress has been made. The approach reported by the researchers includes a 3-axis acceleration and gyro sensor based on the Quaternion calculation method (Tadano et al, 2013), the body inertial sensing network (Guo et al, 2012), an accelerometer and magnetometer (Kun et al, 2011), an anatomical coordinate system using an intra-shoes pressure measurement system (IPM) (Kim and Nussbaum, 2014) and foot switches and micro-controller, connected to an Android Smart-phone using Bluetooth (Wagner and Ganz, 2012).

There are many commercial gait analysis systems based on wearable sensors, but similar devices are still being developed, which is proof of the lack of satisfaction with existing products. Moreover, for individuals in special circumstances, such as patients in the recovery phase of musculoskeletal or nervous system diseases, being able to simultaneously collect the gait analysis results in a free examination environment may provide useful information for measuring and evaluating gait to the treatment provider. In particular, for patients requiring intervention, the collection of spatiotemporal gait parameters without post-processing time is useful to immediately reflect the intervention after measurement (Lopez-Meyer et al, 2011).

Therefore, we proposed a new portable gait analysis system (PGAS) with simple post-processing time, user-friendly, and no experimental site restrictions. The primary goal of this study was to compare the tempo-spatial gait parameters of PGAS developed based on the wearable sensor and the motion capture system to confirm its reliability.


Ⅱ. Methods

1. Subjects

The test-retest study was used to examine the agreement between the same gait parameters such as gait velocity, cadence, step length, stride length, and joint angle (hip, knee, and ankle) obtained with PGAS and MCS. Twenty healthy subjects who had no history of neurological or musculoskeletal pathologies were recruited from the S University. All experimental protocols and procedures were explained to each subject and approved by the institutional review board of Sahmyook University. All subjects provided written informed consent prior to enrollment in this study.

2. Study protocol

A trained examiner explained the tasks through simple instructions before the experiment and leg length (from the great trochanter to the lateral epicondyle, from the lateral epicondyle to the heel through the lateral malleolus) was measured with tape measure. The measurement of the leg length is used for calculating the gait variable by inputting PGAS before the start of the experiment. Participants wore the sensors unit and reflective markers of both devices at the same time without overlapping them, and also walked 12m walkway along the blue line with the ready signals of the measuring administer of both devices. All subjects performed twice time with barefoot and allowed a break time on a chair between trials.

3. Experimental tool

1) Portable Gait Analysis System (PGAS)

The PGAS in this study consisted of eight three-dimensional (3D) inertial measurement unit (IMU) boards as wireless sensors, and one main board as a fusion center. Each IMU board consist of a magnetometer, an accelerometer, a gyroscope, and a Bluetooth transmitter and it can measure the absolute angle of the ankle, shank, thigh, and pelvis based on an axis perpendicular to the ground using a sensor attached to the body segment, and transmit the measured angle data to the main board (Kok et al, 2012). The fusion center receives the raw angle data via the Bluetooth receivers, the MCU of the fusion center combines the angle data of the hip, knee, and ankle, and transmits it to the computer connected by a universal asynchronous receiver and transmitter (UART) and to universal serial bus (USB) bridge.

2) Motion Capture System (MCS)

The motion capture system attaches a marker to a joint point of the human body and analyzes the motion by photographing it with an infrared camera. In this study, we used Qualisys Motion Capture System (Qualisys AB, Savedalen, Sweden, 2012) as the MCS, and the system's ICC (Intra-class Correlation Coefficient) is reported from 0.88 to 0.97 (Ghariba et al, 2011). The camera system was calibrated on every morning prior to experimental and system was tracked the position of reflective markers in the room and calculates the temporal and spatial gait parameters based on foot contact and foot off events with six infrared cameras recoding at 100 Hz. 'Track Manager' (Track Manager version 2.5, Qualisys, Sweden, 2012) processed and calculated temporal and spatial gait parameters of motion and force plate data and Visual 3D Basic' (C-Motion, USA, 2012) was used to digitally reconstruct the subject's lower body anatomy and calculated the angle-change for the test subject’s left and right hip, knee, and ankle joints. 'Visual 3D' was also used to calculate the force plate data (Eigster, 2010).

4. Measurements

The experiment was conducted as a test-retest and while a 12m walk, the subject maintained a static posture for 10 seconds required to correct the accelerometer gravity before walking, and analyzed after removing both the initial 2m for acceleration and the last 2m for deceleration. The spatial-temporal variables measured in the two systems, such as walking speed, cadence, stride length, stride length, and joint angle, were used to compare the reliability of the two instruments.

5. Data analysis

SPSS version 20 statistical software (IBM, Chicago, IL, USA) was used for all statistical analyses. Results are presented as mean ± standard deviation. The Shapiro-Wilk test was used to analyze the normality of the distribution in the clinical and general characteristics of subjects. The test-retest reliability for each test within the PGAS and intra-tester reliability for each test between the PGAS and the MCS for all tempo-spatial gait parameters were estimated using the ICCs. For all test, statistical significance was set at 0.05.


Ⅲ. Results

For general characteristics of subjects, gender, age, height, weight, and leg length were checked <Table 1>.

General characteristics of subjects (N=20)

The ICCs for each test within the PGAS were found to have good test-retest reliability in gait parameters (0.840 to 0.992) and joint ranges in the stance phase (0.907 to 0.988) and swing phase (0.909 to 0.982) <Table 2 and 3>.

Test-retest reliability of the PGAS for gait parameters (N=20)

Test-retest reliability of the PGAS for joint angles of stance and swing phase (N=20)

The ICCs for each test between systems were found to have acceptable intra-tester reliability in gait parameters (0.809 to 0.961) and joint angles in the stance phase (0.800 to 0.977) and swing phase (0.817 to 0.942) <Table 4 and 5>.

ICCs for gait parameters between systems (N=20)

ICCs for joint angles of stance and swing phase between systems (N=20)


Ⅳ. Discussion

We have demonstrated that the reliability for measured tempo-spatial gait parameters such as velocity, cadence, step length, stride length, and angle in joint (ankle, hip, and knee) of healthy subjects, using a newly developed PGAS. As a result, the ICC for each test within the developed PGAS showed strong agreement on measurement repeatability and acceptable reliability compared to the MCS.

In a similar approach, Tadano et al (2013) suggested the implementation of wearable tri-axial acceleration and gyro sensors attached to a segment of the lower limbs, and estimated the angular velocity data during gait in five healthy individuals. The joint trajectory was in the horizontal and sagittal planes, and a comparison of this system was performed with a camera-based motion analysis system, with average root mean square error (RMSE) and correlation coefficient (CC) of 10.14˚ and 0.98 for the hip flexion, 7.88˚ and 0.97 for the knee flexion, and 9.75˚ and 0.78 for the ankle flexion angles. As a result, the RMSE represented a variation; however, with the high CC observed, researchers proposed that this was caused by inaccuracies in calculating the measurements from the camera images.

In this study, the sensor units were constructed with the tri-axial acceleration and gyro sensors as in the previous study; however, it could measure each joint angle and gait parameter such as velocity, cadence, step length, and stride length simultaneously, and showed good ICCs compared with the MCS.

The body inertial-sensing network, which consisted of one three-axis accelerometer, one three-axis magnetometer, and one three-axis gyroscope, was tested by Guo et al (2012) and seven hemiplegia patients and seven healthy subjects were participated but only calculated the angle of knee flexion and extension in one gait cycle and then the researchers were considering other parameters such as the step length or hip joint angle in their future works. One of the advantages of PGAS developed compared to previous studies is that it can detect the sequence of gait cycles from the beginning to the end, and it is more appropriate for the analysis of the entire gait cycle to understand the variability during gait.

Lopez-Meyer et al (2011) used a wearable footwear-based sensor system like a Force-Sensitive Resister (FSR) and accelerometer to describe the temporal gait parameters of seven individuals with stroke and sixteen healthy subjects. This system automatically detected heel-strike and toe-off events, obtained temporal gait parameters as compared to the GAITRite system, and showed no significant difference in the cadence (p>0.35) and temporal gait parameters (p>0.18) in healthy subjects compared with the cadence (p>0.29) and temporal gait parameters (p>0.51) in subjects with stroke. As discussed, this method using FRS sensors only estimated cadence and temporal gait parameters. The present study focused on the accurate estimation of spatial and temporal parameters like velocity, walked distance, and joint angles of a subject by using data calculated from the accelerometer and gyro sensors. However, automatic detection of gait event is one of the strengths of wearable footwear-based sensor system; therefore, we also consider supplementation of FRS sensors in our future study.

Although decisive and excellent reliability were found in both test-retest reliability of a PGAS and intra-tester reliability of the two systems, the ICC for the ankle joint was lower than that for other joints in healthy subjects during gait. This indicated that there was a possibility of errors in the MCS or the PGAS. According to many researchers, even though Vicon motion analysis systems are referred to as the “Gold standard” for 3-D motion tracking system, they also have been reported an accuracy errors (Dorociak and Cuddeford, 1995; Windolf et al, 2008). The gait analysis system using a wearable sensor has also limitations such as attachment problems with strap and external signal noise (Tadano et al, 2013).

Regarding future studies on system implementation, several items could be considered for improvement. First, even though errors are very small, considering both a roll and a pitch can result in more precise stride or step lengths since a pitch can never be zero when patients walk. Second, since it seems that joint angles, especially angles of ankle, change faster than the 100 ms sampling speed, the sampling rate should be increased to, for example, 1 ms or 10 ms by adjusting the spaces of the Bluetooth devices due to the interference among them. Third, mechanical noises have been noticed owing to the body vibration when subjects walk. Therefore, since the phenomenon happens more with bigger or taller devices, smaller IMU devices with small batteries is preferred. Fourth, since the system provides only joint angles and step lengths, researchers use Excel to calculate stride lengths, cadences, velocities, and stance or swing durations. Instead, analytical tools could be embedded in the system, so that researchers can obtain the desired data easily. During experiments, researchers have to recharge the IMU devices from time to time, which means a low power design or high capacity batteries would be beneficial for researchers.


Ⅴ. Conclusion

The results of this study indicated that the developed PGAS showed acceptable agreement compared with the MCS, high test-retest reliability within the PGAS for dynamic gait analysis, and good grade of ICCs compared to the results from the MCS statistically. The gait kinematic parameters in healthy subjects can be used as standard values for using this PGAS.

Acknowledgments

This study was supported by a grant from the NRF (NRF-2018R1D1A1B07045746), which is funded by the Korean government.

References

  • Alandry B, Dumas N, Latorre L, et al. A CMOS multi-sensor system for 3D orientation determination. In 2008 IEEE Computer Society Annual Symposium on VLSI. IEEE; 2008. p.57-62. [https://doi.org/10.1109/ISVLSI.2008.80]
  • Dorociak RD, Cuddeford TJ. Determining 3-D system accuracy for the VICON 370 system. Gait Posture 1995;3(2):88. [https://doi.org/10.1016/0966-6362(95)93468-R]
  • Dobkin BH, Firestine A, West M, et al. Ankle dorsiflexion as an fMRI paradigm to assay motor control for walking during rehabilitation. Neuroimage 2004;23(1):370-81. [https://doi.org/10.1016/j.neuroimage.2004.06.008]
  • Eigster M. Reliability assessment of a Qualisys 3D gait analysis system [Unpublished doctoral dissertation]. University of Iceland; 2010.
  • Gharib NM, Adel SM, Kishk NA. Reliability of three-dimensional motion analysis in assessment of Bell’s palsy. J Am Sci 2011;7(9):126-34.
  • Ghoussayni S, Stevens C, Durham S, et al. Assessment and validation of a simple automated method for the detection of gait events and intervals. Gait Posture 2004;20(3):266-72. [https://doi.org/10.1016/j.gaitpost.2003.10.001]
  • Guo Y, Wu D, Liu G, et al. A low-cost body inertial-sensing network for practical gait discrimination of hemiplegia patients. Telemed J E Health 2012;18(10):748-54. [https://doi.org/10.1089/tmj.2012.0014]
  • Kim S, Nussbaum MA. Evaluation of two approaches for aligning data obtained from a motion capture system and an in-shoe pressure measurement system. Sensors (Basel) 2014;14(9):16994-7007. [https://doi.org/10.3390/s140916994]
  • Kok M, Hol JD, Schön TB, et al. Calibration of a magnetometer in combination with inertial sensors. In 2012 15th International Conference on Information Fusion. IEEE; 2012. p.787-93.
  • Kun L, Inoue Y, Shibata K, et al. Ambulatory estimation of knee-joint kinematics in anatomical coordinate system using accelerometers and magnetometers. IEEE Trans Biomed Eng 2011;58(2):435-42. [https://doi.org/10.1109/TBME.2010.2089454]
  • Lopez-Meyer P, Fulk GD, Sazonov ES. Automatic detection of temporal gait parameters in poststroke individuals. IEEE Trans Inf Technol Biomed 2011;15(4):594-601. [https://doi.org/10.1109/TITB.2011.2112773]
  • Park J, Kim E, Yang S, et al. Effects of Random Visual and Auditory Stimulation on Walking of Healthy Adults. J Korean Phys Ther Sci 2019;26(1):35-44. [https://doi.org/10.26862/jkpts.2019.06.26.1.35]
  • Tadano S, Takeda R, Miyagawa H. Three dimensional gait analysis using wearable acceleration and gyro sensors based on quaternion calculations. Sensors(Basel) 2013;13(7):9321-43. [https://doi.org/10.3390/s130709321]
  • Tao W, Liu T, Zheng R, et al. Gait analysis using wearable sensors. Sensors(Basel) 2012;12(2):2255-2283. [https://doi.org/10.3390/s120202255]
  • Wagner R, Ganz A. PAGAS: Portable and accurate gait analysis system. Conf Proc IEEE Eng Med Biol Soc 2012;280-3. [https://doi.org/10.1109/EMBC.2012.6345924]
  • Windolf M, Götzen N, Morlock M. Systematic accuracy and precision analysis of video motion capturing systems--exemplified on the Vicon-460 system. J Biomech 2008;41(12):2776-80. [https://doi.org/10.1016/j.jbiomech.2008.06.024]
  • Yavuzer G, Oken O, Elhan A, et al. Repeatability of lower limb three-dimensional kinematics in patients with stroke. Gait Posture 2008;27(1):31-5. [https://doi.org/10.1016/j.gaitpost.2006.12.016]

Table 1.

General characteristics of subjects (N=20)

Variables M ± SD
aM±SD
Gender male
Age (years) 26.15±4.68a
Height (cm) 176.15±4.81
Weight (kg) 69.25±8.15
Leg length (Rt) Thigh 41.28±2.59
Shank 49.48±2.61
Leg length (Lt) Thigh 41.35±2.56
Shank 49.68±2.36

Table 2.

Test-retest reliability of the PGAS for gait parameters (N=20)

Test Reset ICC 95% confidence interval
MIN MAX
aM±SD, PGAS=portable gait analysis system; Rt=right; Lt=left
Velocity (m/sec) 1.081±0.208a 1.064±0.215 0.973 0.931 0.989
Cadence (step/min) 104±13.790 105±14.511 0.992 0.970 0.997
Rt step length (m) 0.612±0.064 0.612±0.060 0.975 0.936 0.990
Lt. step length (m) 0.615±0.064 0.603±0.056 0.910 0.772 0.964
Stride length (m) 1.216±0.114 1.227±0.114 0.974 0.933 0.99
Rt. stance (%) 60.633±3.023 61.000±2.256 0.840 0.597 0.937
Lt. swing (%) 39.367±3.023 38.999±2.258 0.841 0.597 0.937
Rt. stance (%) 60.988±2.569 60.596±2.791 0.962 0.903 0.985
Lt. swing (%) 39.012±2.569 39.403±2.791 0.962 0.903 0.985

Table 3.

Test-retest reliability of the PGAS for joint angles of stance and swing phase (N=20)

TEST RESET ICC 95% confidence interval
MIN MAX
aM±SD, PGAS=portable gait analysis system; MCS=motion capture system; Rt=right; Lt=left
Stance phase Rt hip (degree) 36.882±4.091a 37.114±3.997 0.988 0.970 0.995
Lt hip (degree) 37.687±4.271 37.219±4.271 0.972 0.929 0.989
Rt knee (degree) 34.439±2.980 35.273±2.567 0.962 0.904 0.985
Lt knee (degree) 33.384±2.977 33.966±3.280 0.960 0.899 0.984
Rt ankle (degree) 21.672±2.538 21.382±1.950 0.918 0.792 0.967
Lt ankle (degree) 22.985±2.071 21.672±2.587 0.907 0.765 0.963
Swing phase Rt hip (degree) 32.773±4.826 33.265±4.398 0.982 0.954 0.993
Lt hip (degree) 31.624±4.876 33.394±4.647 0.924 0.809 0.970
Rt knee (degree) 57.468±3.671 57.926±3.191 0.946 0.864 0.979
Lt knee (degree) 57.043±4.713 57.353±3.336 0.909 0.771 0.964
Rt ankle (degree) 18.868±3.393 19.190±2.714 0.953 0.882 0.981
Lt ankle (degree) 19.688±4.624 20.058±3.928 0.943 0.857 0.978

Table 4.

ICCs for gait parameters between systems (N=20)

PGAS MCS ICC 95% confidence interval
MIN MAX
aM±SD, PGAS=portable gait analysis system; MCS=motion capture system; Rt=right; Lt=left
Velocity (m/sec) Test 1.081±0.209a 1.057±0.134 0.911 0.776 0.965
Retest 1.064±0.215 1.076±0.179 0.886 0.712 0.955
Cadence (step/min) Test 104.833±13.790 102.010±10.417 0.869 0.67 0.948
Retest 105.444±14.511 104.17±9.181 0.833 0.577 0.934
Rt step length (m) Test 0.612±0.060 0.614±0.065 0.842 0.601 0.937
Retest 0.612±0.064 0.613±0.066 0.884 0.708 0.954
Lt. step length (m) Test 0.603±0.056 0.620±0.059 0.809 0.517 0.924
Retest 0.615±0.063 0.619±0.058 0.876 0.687 0.951
Stride length (m) Test 1.215±0.114 1.234±0.121 0.861 0.649 0.945
Retest 1.227±0.123 1.233±0.111 0.939 0.846 0.976
Rt. stance (%) Test 60.633±3.023 60.259±2.800 0.811 0.522 0.925
Retest 61.000±2.256 61.371±2.258 0.812 0.526 0.926
Lt. swing (%) Test 39.367±3.023 39.741±2.800 0.811 0.522 0.925
Retest 38.999±2.257 38.628±2.258 0.812 0.526 0.926
Rt. stance (%) Test 60.988±2.569 61.377±2.745 0.961 0.901 0.999
Retest 60.596±2.791 61.411±3.639 0.906 0.763 0.963
Lt. swing (%) Test 39.012±2.569 38.623±2.745 0.961 0.901 0.985
Retest 39.404±2.791 38.589±3.639 0.906 0.763 0.963

Table 5.

ICCs for joint angles of stance and swing phase between systems (N=20)

PGAS MCS ICC 95% confidence interval
MIN MAX
aM±SD, PGAS=portable gait analysis system; MCS=motion capture system; Rt=right; Lt=left
Stance phase Rt hip (degree) Test 36.883±4.091a 36.379±4.068 0.977 0.941 0.991
Retest 37.113±3.997 37.634±3.281 0.943 0.855 0.977
Lt hip (degree) Test 37.687±3.489 36.679±3.990 0.890 0.721 0.956
Retest 37.219±4.271 36.679±3.772 0.836 0.585 0.935
Rt knee (degree) Test 34.439±2.980 33.450±3.610 0.821 0.548 0.929
Retest 35.273±2.567 34.324±3.664 0.842 0.600 0.937
Lt knee (degree) Test 33.384±2.977 34.388±2.553 0.888 0.718 0.956
Retest 33.966±3.280 35.525±2.710 0.899 0.744 0.960
Rt ankle (degree) Test 21.672±2.538 22.368±2.154 0.833 0.578 0.934
Retest 21.382±1.950 22.154±2.261 0.856 0.635 0.943
Lt ankle (degree) Test 22.985±2.071 21.492±1.762 0.870 0.671 0.948
Retest 21.672±2.587 22.576±1.784 0.800 0.495 0.921
Swing phase Rt hip (degree) Test 32.773±4.826 31.999±5.593 0.942 0.853 0.977
Retest 33.265±4.398 31.874±4.266 0.863 0.654 0.946
Lt hip (degree) Test 31.624±4.876 32.483±6.129 0.884 0.708 0.954
Retest 33.394±4.647 31.243±4.609 0.871 0.675 0.949
Rt knee (degree) Test 57.468±3.671 57.430±3.350 0.890 0.721 0.956
Retest 57.926±3.191 58.062±4.398 0.835 0.583 0.935
Lt knee (degree) Test 57.043±56.252 56.252±4.915 0.890 0.722 0.956
Retest 57.353±3.336 57.314±5.927 0.863 0.655 0.946
Rt ankle (degree) Test 18.868±3.393 19.259±3.909 0.872 0.676 0.949
Retest 19.190±2.714 18.069±3.495 0.933 0.831 0.973
Lt ankle (degree) Test 19.688±4.624 19.594±4.667 0.939 0.846 0.976
Retest 20.058±3.928 19.433±4.695 0.817 0.537 0.927