[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Article in Journal
Deep Learning-Based Adaptive Neural-Fuzzy Structure Scheme for Bearing Fault Pattern Recognition and Crack Size Identification
Next Article in Special Issue
Advances in Therapeutic Monitoring of Lithium in the Management of Bipolar Disorder
Previous Article in Journal
Human Body-Related Disease Diagnosis Systems Using CMOS Image Sensors: A Systematic Review
Previous Article in Special Issue
Dynamic Hand Gesture Recognition Using 3DCNN and LSTM with FSM Context-Aware Model
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Sensors
Volume 21
Issue 6
10.3390/s21062104
sensors-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Kinect V2-Based Gait Analysis for Children with Cerebral Palsy: Validity and Reliability of Spatial Margin of Stability and Spatiotemporal Variables

by
Yunru Ma
1,
Kumar Mithraratne
2,
Nichola Wilson
3,
Yanxin Zhang
1,* and
Xiangbin Wang
4,5,*
1
Department of Exercise Sciences, The University of Auckland, Auckland 1142, New Zealand
2
Auckland Bioengineering Institute, The University of Auckland, Auckland 1142, New Zealand
3
Department of Surgery, The University of Auckland, Auckland 1142, New Zealand
4
College of Rehabilitation Medicine, Fujian University of Traditional Chinese Medicine, Fuzhou 350122, China
5
Key Laboratory of Orthopedics and Traumatology of Traditional Chinese Medicine and Rehabilitation, Fujian University of Traditional Chinese Medicine, Ministry of Education, Fuzhou 350122, China
*
Authors to whom correspondence should be addressed.
Sensors 2021, 21(6), 2104; https://doi.org/10.3390/s21062104
Submission received: 16 December 2020 / Revised: 15 January 2021 / Accepted: 13 March 2021 / Published: 17 March 2021
(This article belongs to the Collection Sensors for Globalized Healthy Living and Wellbeing)
Browse Figures
Figure 1
<p>Lower limb reflective marker locations of a modified Cleveland Clinic marker set (<b>a</b>) and joint landmarks recorded by the Kinect V2 sensor (<b>b</b>).</p> "> Figure 2
<p>Illustration of the calculation of margin of stability (MOS).</p> "> Figure 3
<p>Bland–Altman plots indicate the agreement between MOS variables for ten children with cerebral palsy (CP) calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.</p> "> Figure 4
<p>Bland–Altman plots indicate the agreement between gait spatiotemporal variables for ten children with CP calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.</p> ">
Versions Notes

Abstract

:
Children with cerebral palsy (CP) have high risks of falling. It is necessary to evaluate gait stability for children with CP. In comparison to traditional motion capture techniques, the Kinect has the potential to be utilised as a cost-effective gait stability assessment tool, ensuring frequent and uninterrupted gait monitoring. To evaluate the validity and reliability of this measurement, in this study, ten children with CP performed two testing sessions, of which gait data were recorded by a Kinect V2 sensor and a referential Motion Analysis system. The margin of stability (MOS) and gait spatiotemporal metrics were examined. For the spatiotemporal parameters, intraclass correlation coefficient (ICC2,k) values were from 0.83 to 0.99 between two devices and from 0.78 to 0.88 between two testing sessions. For the MOS outcomes, ICC2,k values ranged from 0.42 to 0.99 between two devices and 0.28 to 0.69 between two test sessions. The Kinect V2 was able to provide valid and reliable spatiotemporal gait parameters, and it could also offer accurate outcome measures for the minimum MOS. The reliability of the Kinect V2 when assessing time-specific MOS variables was limited. The Kinect V2 shows the potential to be used as a cost-effective tool for CP gait stability assessment.

1. Introduction

Postural control is the ability to control the body’s position in space to achieve stability, requiring the centre of mass (COM) to be positioned within the boundary of the base of support (BOS) [1]. In the case of cerebral palsy (CP), deteriorations and delays in motor skill acquisition and development are associated with weak postural control schema, which is a significant component of gait disorders [2,3]. Children with CP are prone to fall. The fall rate of inpatient children with CP was reported to be 27% in a prospective study [4]. Approximately 35% of children reported daily falling, and 30% of them fell monthly or weekly according to a retrospective study [5]. Although most falls cause few consequences with no more than cuts and bruises, these nonfatal injuries could lead to a significant healthcare burden with increased pain, disability, length of stay, absence for treatment, and financial expenses [6].
About 50% of the falls happen during walking [7]. Children with CP tend to walk inefficiently with a larger gait velocity and stride length and demonstrate an inappropriate coronal foot placement [8]. Besides, they generate more cadences, trunk compensations, and conservative lateral postural control strategies to reestablish stability when encountering perturbances during gait [9,10,11]. Improved walking ability, such as less proneness to falls, positively affects the accomplishment of life habits and social integration [12,13]. Therefore, optimising walking ability and mobility is one of the therapeutic goals for patients with CP. Various rehabilitation programmes have been implemented to achieve this goal. For example, substantial evidence has proven that functional gait training that involves diverse interventions has clinically significant benefits on children and young adults with CP [14]. To obtain in-depth knowledge about the effect of gait rehabilitation programmes on patients with CP, quantitatively measuring gait characteristics is needed [15]. Besides, quantitative aspects of gait can directly reflect the impaired postural control ability that influences gait stability [16]. The relationship between gait quantitative features and fall risks can facilitate gait monitoring and help researchers develop targeted intervention programmes [17]. Some spatiotemporal gait outcomes, such as gait speed, walking endurance, and step length, are widely utilised among studies to assess gait stability and evaluate treatment outcomes [14,15,18,19,20].
The margin of stability (MOS) is also one of the quantitative gait measurements that evaluate the COM’s motion relative to the BOS that determined by the foot placement [21]. The MOS calculation is based on the inverted pendulum model representing the human body’s upright posture [22]. According to Hof et al.’s method, MOS is calculated as the distance between the edge of BOS and the vertical projection of the extrapolated COM, which is defined as the velocity of COM relative to the height of COM over BOS [21]. An individual is considered unstable or with a high risk of falling when the extrapolated COM exceeds BOS boundaries (e.g., MOS is negative). From the biomechanical perspective, MOS represents the “initial condition” when the subject encounters a perturbation. Compared with other stability assessment approaches that only quantify average measures, MOS is able to evaluate dynamic stability at a specific time step during locomotion [9,23].
With this advantage, MOS has already been widely applied to evaluate gait abnormality and fall risks in a wide range of clinical populations, including patients with stroke [24], Parkinson’s Disease [25], hereditary spastic paraparesis [26], as well as CP [9,27]. In a recent prospective observational study, Mehdizadeh et al. found that minimum mediolateral MOS was the most important hallmark when assessing the short-term fall risk for patients with dementia [16]. A retrospective study demonstrated that MOS was significantly correlated with the fall history of patients with multiple sclerosis, and it could be considered as an effective measure to assess dynamic stability during gait [28].
Generally, parameters required for MOS computation are recorded by optoelectronic motion capture systems combined with force plates. However, these high-end infrared cameras are costly and unfriendly to nonspecialists. Its marker-based tracking strategy may cause unnatural performance for children with CP and make marker placement a challenging job. Therefore, there is a growing need to develop cost-effective and portable gait analysis systems for clinic-based gait progression monitoring and home-based gait investigation, which can empower patients to take an active role in their health management, provide convenient and accessible options for people living in remote areas [29], and also ensure uninterruptable longitudinal self-monitoring during the epidemic time.
The Microsoft Kinect sensor, which was initially designed as a motion-sensing input device for interactive games, can perform real-time motion capture without any passive markers or handheld controllers. Compared to other innovative techniques, such as activity-tracking devices or wearable accelerometer-based devices, the Kinect has the advantage of providing unobtrusive investigation [30]. Previous studies have proposed Kinect-based gait stability assessment systems to monitor fall risks for geriatric populations [16,31]. However, very little is known about the Kinect’s validity for estimating MOS for a pediatric group like CP. Considering the relative accurate spatiotemporal tracking accuracy [32,33], it seems that MOS variables could be accurately estimated due to their spatiotemporal-related features. Since clinical measures should be valid, reliable, and responsive to meaningful changes when used with relation to interventions [13], it is logical and necessary to determine the validity and reliability of MOS derived from the Kinect before it could be used in real clinical scenarios. From the clinical aspect, the same outcome measure’s validity and reliability could differ among different patient groups [13,34]. Although the accuracy of gait spatiotemporal parameters estimated by the Kinect has been examined among adults with or without abnormal gait patterns [32,35,36], evidence involving the CP population is very limited [37]. Moreover, many more studies only covered the Kinect’s validity rather than both validity and reliability, which may hamper its application in real practice.
To fill the gaps mentioned above, we proposed a method to compute MOS parameters from skeleton landmarks tracked by the Kinect V2 sensor. The results were compared against data measured by a referential optoelectronic Motion Analysis system. This study aimed to assess the concurrent validity and interday reliability of the Kinect V2 when assessing MOS and spatiotemporal parameters of overground gait for children with CP. We hypothesised that the Kinect V2 could provide accurate and reliable measurements for MOS and spatiotemporal gait parameters.
The rest of the paper is as follows: “Methods” section details the characteristics of ten CP participants, the definition of selected gait spatiotemporal parameters, the method of MOS calculation, as well as statistical methods implemented to determine the validity and reliability of the Kinect V2 sensor; “Results” section describes the gait spatiotemporal and MOS parameters estimated by per device on per testing day. The between-device and between-testing day comparison is given in this section. “Discussion” section discusses the findings that arose from the results, analyses possible reasons that lead to the between-device and between-testing day deviations, and discuss the Kinect’s potential implementation in gait stability management for patients with CP, followed by the conclusive remarks in the “Conclusion” section.

2. Methods

2.1. Participants

Ten children with CP (three males and seven females, age = 6.4 ± 2.2 years, body mass = 23.2 ± 7.4 kg, body height = 116.7 ± 11.0 cm) were recruited and presented for the gait data collection. Participants were diagnosed as hemiplegia (one child), diplegia (five children), quadriplegia (two children), or dyskinesia (two children), and they were classified as Level I (three children) or II (seven children) on the Gross Motor Function Classification System (GMFCS) [38]. Children who had a significant illness, injury, or surgery within the previous six months that may have impacted their usual activity levels in the community, or was not possible to complete a three-dimentional gait analysis, were excluded from this study. Ethics approval was obtained from the local institutional review board and the University of Auckland Human Participants Ethics Committee. Written consent was obtained from each child’s parent or guardian along with assent from the child. All of the ten children were able to independently undergo gait analysis without any assistance from other people or walking aids.

2.2. Data Collection

As illustrated in Figure 1a, reflective markers were placed on the participants’ lower limb according to a modified Cleveland Clinic marker set [39,40]. This markerset consisted of the sacrum, right/left anterior superior iliac spine (RASI or LASI), rigid thigh clusters of markers (R.Thigh or L.Thigh), lateral and medial condyles of the knee (RKNE or RKNM or LKNE or LKNM), rigid shank clusters of markers (R.Shank/L.Shank), lateral and medial malleoli of the ankle (RANK or RANM or LANK or LANM), calcaneus (RHEE or LHEE), and the second metatarsal (RTOE or LTOE) of both feet.
An eight-infrared-camera motion capture system (Motion Analysis Corporation, Santa Rosa, CA, USA) recorded the reflective marker trajectories at a sampling rate of 100 Hz. The length of the calibrated motion capture volume was about 4 m. The Kinect V2 sensor (Microsoft Cop., Redmond, WA, USA) was placed on a tripod with a height of 0.8 m and a distance of 5 m from the start line of the walkway. The Kinect V2 sensor was placed in front of the participant to provide a frontal view, ensuring the tracking accuracy for both kinematic and spatiotemporal parameters. It was triggered simultaneously with the Motion Analysis system, recording coordinates of twenty-five skeletal landmarks (Figure 1b) at a fluctuating frequency around 30 Hz through a custom-written software program.
Three-dimensional gait analysis test consisted of two testing sessions. In the first testing session, participants were given sufficient time to try to walk along the walkway. After they became familiar with the testing procedures, they were asked to performed overground gait trials at a self-selected speed along the level walkway independently. Children walked from the starting line towards the Kinect V2 sensor barefoot. Each participant completed at least three successful gait trials. A successful gait trial should contain at least one complete right gait cycle with all the landmarks, and reflective markers could be seen in the motion capture interfaces of two systems. To test the interday reliability of the Kinect V2, participants were asked to attend another gait test session the following day. All the ten participants presented for the Kinect V2-based gait analysis in the second testing session. The experimental setup of the Kinect V2 for the second session was consistent with the previous one. At least three successful gait trials were obtained for each child. The criteria of a successful gait trial were consistent with what was given in the previous session.

2.3. Data Analysis

The data collected by the Kinect V2 and Motion Analysis system were filtered via a second-order Butterworth low-pass digital filter with a cut-off frequency of 6 Hz. Gait events were identified based on the distance between the sacrum and foot landmarks [41,42,43,44]. For the Motion Analysis system, the foot strike (FS) was identified as when the anterior-posterior (AP) distance between the heel of the leading foot and sacrum markers reached the maximum, and the toe-off (TO) was defined as when the AP distance between the second metatarsal marker of the rearfoot and the sacrum reached the maximum. For the Kinect V2, FS and TO events were detected similarly with the Motion Analysis, in which the “spine base” landmark was used as the sacrum, and the ankle landmark was used to represent the heel and second metatarsal due to its better tracking performance than the foot landmark. Spatiotemporal gait parameters were identified according to previous studies [35,45,46]. Definitions of selected spatiotemporal parameters in the Kinect V2 and Motion Analysis systems are presented in Table 1, and their calculation formulas are demonstrated in Table 2.
The COM position was calculated as the average position among the sacrum, LASI, and RASI markers in the Motion Analysis [23,47]. The “spine_base” landmark position was considered as the COM in the Kinect V2 system (Figure 1b). The definition of MOS was presented in Figure 2.
Equations (7)–(9) were used to compute the dynamic MOS based on prior studies [21,23]:
ω 0   =   g l
where ω 0 represented the eigenfrequency of the pendulum, g   =   9.81   m / s 2 was the gravitational constant, and the l was the length of the pendulum, which was defined as the distance between COM and the heel marker (in Motion Analysis) or the ankle landmark (in the Kinect) at the foot strike.
X C O M   =   a + v ω 0
where the X C O M represented the extrapolated COM, a represented positions of COM in the AP or mediolateral (ML) direction. v was the velocity of COM, which was acquired by calculating the first time derivative of COM positions in the AP or ML direction.
M O S   =   B O S − X C O M
M O S was finally obtained as the distance between X C O M and the boundary of B O S in the AP or the ML direction. B O S in the AP direction was identified using the second metatarsal marker in the AP direction of the leading foot, and B O S in the ML direction was estimated using the lateral malleoli marker in the ML direction of the leading foot [24]. M O S at foot strike, minimum M O S during the stance phase and M O S at midstance (the point when the swing limb passed the stance limb in the direction of progression) [48] were extracted for statistical analysis. All the gait event detections, spatiotemporal gait parameters, and M O S computations were conducted in customised scripts written in the Matlab R2019a (MathWorks Inc., Natick, MA, USA).

2.4. Statistics

In each gait trial, one right gait cycle was extracted for analysis. All the MOS and spatiotemporal parameters measured by each device on each testing day were averaged among three gait trials for each participant. The intraclass correlation coefficient (ICC) was used to assess the agreement between the Kinect V2 and Motion Analysis system and the agreement of the Kinect V2 between two testing days. ICCs were estimated, and their 95% confidential intervals were calculated based on a mean rating (k = 3), absolute-agreement, 2-way random-effects model. ICC values were interpreted as: excellent (0.75–1), modest (0.4–0.74), or poor (0–0.39) [49]. The standard error of measurement (SEM) was defined according to Equation (10):
e   =   s × 1 − r
where e was SEM, s was the standard deviation of measurements determined from ANOVA [50,51], and r was ICC.
The relative error (in percentage) was computed to express the absolute between-device measurement difference as a percentage of the measure estimated by the referential Motion Analysis system (Equation (11)) [36].
η   =   ε M o t i o n   A n a l y s i s − ε K i n e c t ε M o t i o n   A n a l y s i s × 100 %
where η stood for the relative error (in percentage), ε M o t i o n   A n a l y s i s and ε K i n e c t were gait parameters measured by Motion Analysis and Kinect V2, respectively.
Furthermore, a Bland–Altman analysis of agreement [52] was performed between gait parameters obtained by the two devices. All the statistical analysis was conducted by using the SPSS statistical package version 25 (SPSS Inc., Chicago, IL, USA).

3. Results

Mean ± 1SD values for each spatial MOS and spatiotemporal gait parameters estimated by each device, inter-device ICC2,k with 95% confidence intervals for agreement, SEM, and relative error (in percentage) are presented in Table 3. The Blan–Altman plots for every selected parameter are presented in Figure 3 and Figure 4; mean difference, LoA, upper and lower LoA are given in Table 4. Mean ± 1SD values for each spatial MOS and spatiotemporal gait parameters acquired by the Kinect V2 on two testing days, interday ICC2,k with 95% confidence intervals for agreement and SEM are presented in Table 5.
For the MOS, an excellent agreement was observed for the minimum MOS during stance in both ML (ICC2,k = 0.81) and AP (ICC2,k = 0.99) directions, and a modest agreement (ICC2,k = 0.42–0.68) was found for MOS at foot strike and midstance in both ML and AP directions between Motion Analysis and the Kinect V2. All the spatiotemporal gait parameters showed excellent agreement between the two devices (ICC2,k = 0.83–0.99). A modest agreement was found in the equivalent pendulum length between the two devices (ICC2,k = 0.45). Mean relative errors (in percentage) ranged from 1.75% to 186.4% for all the selected parameters. MOS at midstance and foot strike in the AP direction demonstrated the most deficient accuracy, with mean relative errors of 112.82% and 186.40%, respectively. All the selected spatiotemporal parameters exhibited lower relative errors (1.75–25.97%).
The interday reliability of most MOS variables was modest (ICC2,k = 0.56–0.69) for the Kinect V2 except for the MOS at midstance in the ML direction, which showed low reliability between two test days (ICC2,k = 0.28). All the spatiotemporal gait parameters (ICC2,k = 0.78–0.88) and equivalent pendulum length (ICC2,k = 0.92) showed excellent reliability between the two testing day.

4. Discussion

This study aimed to assess the validity and reliability of the Kinect V2 when using it to evaluate spatial MOS and spatiotemporal gait parameters during overground walking for children with CP. The results proved that the Kinect V2 could be employed as a valid and reliable screening tool to investigate spatiotemporal gait parameters and their progressive changes. For the MOS evaluation, the Kinect V2 could accurately assess the minimum MOS during the stance phase. These results partially supported our hypothesis. The Kinect V2′s overall moderate to poor interday reliability for MOS evaluation made it hard to distinguish the sources that caused the interday changes in MOS. Its reliability in MOS evaluation should be further verified before it can be widely adopted for in-clinic or home-based gait monitoring.
In this study, spatiotemporal variables derived from the Kinect V2 sensor showed an excellent agreement with their referential counterparts (ICC2,k = 0.83–0.99). In accordance with the present results, previous studies have reported a reasonable validity of the Kinect V2 when assessing spatiotemporal gait parameters regardless of what the walking condition (overground or treadmill-based), walk speed (at a self-comfortable or specific speed), population (healthy or patients with movement disorders), and also the referential motion capture system (optoelectronic cameras, videos, or press sensing mat) was [35,36,44,53,54]. This study also demonstrated good reliability of the Kinect V2 sensor for spatiotemporal measurements (ICC2,k = 0.78–0.88), favouring prior findings [35,55,56,57,58,59]. This study further enhanced the Kinect V2′s feasibility to be used as a valid and reliable tool to screen and investigate ongoing gait spatiotemporal progression for a pediatric population like children with CP.
However, except for the minimum MOS (ICC2,k = 0.81–0.99), other MOS variables did not show such a strong agreement with those obtained by the referential motion capture system (ICC2,k = 0.42–0.68). It was potentially attributed to two main reasons. Firstly, the length of the pendulum was calculated as the distance between COM and foot ground contact in a marker-based motion capture system [23]. In the Kinect V2-based gait analysis system, the ankle landmark was used to calculate MOS parameters instead of the foot centre due to its relatively stable tracking performance, which possibly resulted in a shorter pendulum length and affected the MOS estimation. Therefore, a moderate agreement was found between the length of pendulum calculated by the two motion capture devices (ICC2,k = 0.45). Additionally, using the ankle landmark in MOS computation has the risk of underestimating BOS. A prior study reported that 66% of 492 patients with CP were affected by the intoeing abnormality, and around 25% of them had the out-toeing problem [60]. It indicates that the patients’ toes position medially or laterally with respect to the heel, leading to changes BOS boundaries. However, using the ankle landmark as an alternative to represent the BOS may fail to detect the foot progression during gait, leading to the interdevice deviations in MOS measurement.
Secondly, the MOS at foot strike and midstance were parameters that relied more heavily on the precise identification of related gait events. The Kinect was reported to show an error with ± 2 frames when detecting the heel strike and toe-off event for treadmill walking [41]. Its relatively low and fluctuating sampling frequency indicates that small deviations could lead to large gait detection errors. Two literature review studies reported that the Kinect’s automated body tracking algorithm was exceptionally good for some spatial gait parameters such as step length, width, and asymmetry when assessing the overground gait for young and old adult participants [32,33]. The difficulties in detecting gait events and relatively low and inconsistent sampling frequency hamper the precision of evaluating timing-related gait parameters, therefore giving the possibility of influencing the extraction of time-specific MOS parameters. Except for using the distance between the sacrum and ankle landmarks as the gait event detection criterion, previous studies have proposed various approaches to identify gait events, for example, utilising ankle or knee displacements that derived from depth images [54,61,62], extracting gait events based on some template models [63], identifying gait events according to the velocity of a specific landmark [64], and so on. Latorre et al. compared five gait event detection methods and found that the validity level varied among the selected methods for both healthy participants and stroke survivors when measuring spatiotemporal gait parameters [36]. Since MOS at midstance and mid-swing are significant indices for walking stability assessment [9,28], there is sufficient room for further progress in determining which gait event definition is the most appropriate one for a specific clinical group, such as children with CP, before the Kinect could be extrapolated to clinical applications.
Although the Kinect V2 could provide reliable interday spatiotemporal evaluation, it seems that the changes in clothing, accommodation to the testing procedure, and the patients’ performance over repeated gait trials may potentially result in the interday deviations in the MOS calculation (ICC2,k = 0.28–0.69). De Jong et al. found that the test-retest ICCs of MOS are lower in healthy participants than patients with balance problems [65]. Since the ICC is determined by both between-subject variability and test-retest variability, a smaller between-subject variability in the healthy control group and similar test-retest variability may collectively result in lower ICCs [65]. In this study, the patient group was less homogeneous. Therefore, it could be induced that a comparable smaller interday variability may lead to lower ICCs in this study. Children in GMFCS Levels II and III exhibit lower within-session variability in their gait patterns when compared to children in GMFCS Level I [66]. Most of the participants of this study were classified in GMFCS Levels II, implying a higher variability of their gait between the two testing days. The MOS calculation seems to be sensitive to interday variabilities, making it hard to distinguish the sources resulting in changes in MOS. The Kinect V2 could provide accurate minimum MOS values (ICC2,k = 0.81–0.99), which are essential gait metrics to assess the risk of falls [16]. However, its reliability of evaluating MOS should be further verified before utilising as a low-cost option to monitor fall risks during gait for children with CP.
From the perspective of clinical application, although the Kinect V2 is unable to provide precise and robust MOS estimation, it still has the potential to be utilised as a promising alternative tool for CP gait stability management. Firstly, the validity and reliability of Kinect V2-based gait spatiotemporal metrics are satisfactory. A previous systematic review study has investigated that some spatiotemporal parameters (i.e., double limb support time, step length, stride length, and step width) are useful features to distinguish CP gait stability from their typically developing peers [15]. Moreover, the gait speed is an essential index to evaluate the effectiveness of gait rehabilitation programmes for patients with CP [14,20]. Thus, the feasibility of the Kinect V2 to be applied as a cost-effective tool for assessing and monitoring gait stability for patients with CP has preliminarily warranted. Secondly, it is known that virtual reality (VR)-assisted gait training programmes enhance treatment outcomes by providing strong motivation and increasing concentration for participants [14]. Considering impaired sensory feedback networks associated with CP, multisensory feedback of an executed movement given by VR can furnish participants with enriching knowledge of performance [14,67]. Since the Kinect is advantageous in real-time motion sensing, it can be involved as an input device in VR-assisted gait stability training systems. Meanwhile, patients can learn and modify their postural control strategies with the help of the Kinect-based real-time gait stability assessment.
Except for the small sample size, some other limitations were encountered in this study. Firstly, to simplify the MOS calculation process, the “spine base” marker was used to represent the COM. However, the COM position with respect to the segmental coordinate system is generally predicted via regression equations with a full-body marker set in traditional optoelectronic motion capture systems [68]. It was found that COM positions decided by 7–9 multisegmental models allow a compromise between feasibility and accuracy when assessing the dynamic stability [69]. Similar anthropometric models and statistically equivalent serial chain methods have been developed to improve the COM estimation of the Kinect [70]. Further research should be undertaken to investigate the influence of different COM computation methods in COM displacement and MOS evaluation. Secondly, only one Kinect sensor was placed in front of the participants in this study to ensure the accuracy of both spatiotemporal and kinematic estimations. The single Kinect V2 setting has very limited capture volume, possibly resulting in missing data when falls occur outside the sensor’s field-of-view in a clinic or home-based environment. The application of multiple Kinect sensors presented a promising prospect to enlarge the volume for motion capture and solve occlusion problems [71]. Meanwhile, depth data obtained from the Kinect could be incorporated with other motion-sensing information, such as data recorded by accelerometers [72] and inertial sensors [73], to develop a multimodel gait stability assessment system for clinic or home-based monitoring. Thirdly, this study used ankle landmarks instead of foot landmarks to represent the BOS because the foot tracking is usually noisy and inaccurate in the Kinect V2. In contrast, the newest Azure Kinect sensor released in 2019 demonstrates significantly better foot tracking accuracy and more precise gait spatiotemporal parameter assessment, indicating improved image sensing strategies [74]. Therefore, in future investigations, it may be possible to use multiple Azure Kinect sensors to develop the precision of gait detection and skeleton tracking performance.

5. Conclusions

Results of this study reveal a reasonable validity and reliability level in evaluating spatiotemporal parameters and, in contrast, less strong validity and reliability when calculating MOS related variables, especially MOS at foot strike and midstance. The Kinect V2 was only firmly valid to provide minimum MOS measures, which is an essential gait metric to track the risk of falls [16]. These results comprehensively suggest that the Kinect V2 could be utilised as a low-cost, portable gait assessment tool to screen and observe ongoing gait progression for children with CP, especially for the spatiotemporal aspect. Future studies should involve more sensors and robust algorithms to enhance skeleton tracking, gait event detection performance, and MOS computation.

Author Contributions

Conceptualisation, Y.Z., K.M. and N.W.; methodology, Y.Z. and Y.M.; data collection and analysis, Y.M.; resources, X.W.; writing—original draft preparation, Y.M.; writing—review and editing, K.M., N.W., Y.Z. and X.W. All authors have read and agreed to the published version of the manuscript.

Funding

Y.M. was funded by China Scholarship Council, grant number 201606520001. This research was supported by the National Natural Science Foundation of China, grant number 82074515.

Institutional Review Board Statement

The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of The University of Auckland (protocol code 019410, date: 18 August 2017).

Informed Consent Statement

Informed consent was obtained from all subjects involved in this study.

Data Availability Statement

The data presented in this study are available on request from the corresponding authors.

Acknowledgments

Thank you to the children and their families for their support in this study.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Pavão, S.L.; dos Santos, A.N.; Woollacott, M.H.; Rocha, N.A.C.F. Assessment of postural control in children with cerebral palsy: A review. Res. Dev. Disabil. 2013, 34, 1367–1375. [Google Scholar] [CrossRef] [Green Version]
  2. Donker, S.F.; Ledebt, A.; Roerdink, M.; Savelsbergh, G.J.P.; Beek, P.J. Children with cerebral palsy exhibit greater and more regular postural sway than typically developing children. Exp. Brain Res. 2008, 184, 363–370. [Google Scholar] [CrossRef] [Green Version]
  3. Rose, J.; Wolff, D.R.; Jones, V.K.; Bloch, D.A.; Oehlert, J.W.; Gamble, J.G. Postural balance in children with cerebral palsy. Dev. Med. Child Neurol. 2002, 44, 58–63. [Google Scholar] [CrossRef]
  4. Alemdaroğlu, E.; Özbudak, S.D.; Mandiroğlu, S.; Biçer, S.A.; Özgirgin, N.; Uçan, H. Predictive Factors for Inpatient Falls among Children with Cerebral Palsy. J. Pediatric Nurs. 2017, 32, 25–31. [Google Scholar] [CrossRef] [PubMed]
  5. Boyer, E.R.; Patterson, A.A. Gait pathology subtypes are not associated with self-reported fall frequency in children with cerebral palsy. Gait Posture 2018, 63, 189–194. [Google Scholar] [CrossRef]
  6. Craig, F.; Castelnuovo, R.; Pacifico, R.; Leo, R.; Trabacca, A. Falls in Hospitalized Children With Neurodevelopmental Conditions: A Cross-sectional, Correlational Study. Rehabil. Nurs. J. 2018, 43, 335–342. [Google Scholar] [CrossRef]
  7. Morgan, P.; McDonald, R.; McGinley, J. Perceived Cause, Environmental Factors, and Consequences of Falls in Adults with Cerebral Palsy: A Preliminary Mixed Methods Study. Rehabil. Res. Pract. 2015, 2015, 196395. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  8. Chang, C.-F.; Wang, T.-M.; Lo, W.-C.; Lu, T.-W.; Hong, S.-W.; Huang, C.-H.; Shieh, J.-Y.; Huang, S.-C. Balance control during level walking in children with spastic diplegic cerebral palsy. Biomed. Eng. Appl. Basis Commun. 2011, 23, 509–517. [Google Scholar] [CrossRef]
  9. Tracy, J.B.; Petersen, D.A.; Pigman, J.; Conner, B.C.; Wright, H.G.; Modlesky, C.M.; Miller, F.; Johnson, C.L.; Crenshaw, J.R. Dynamic stability during walking in children with and without cerebral palsy. Gait Posture 2019, 72, 182–187. [Google Scholar] [CrossRef]
  10. Kurz, M.J.; Arpin, D.J.; Corr, B. Differences in the dynamic gait stability of children with cerebral palsy and typically developing children. Gait Posture 2012, 36, 600–604. [Google Scholar] [CrossRef]
  11. Iosa, M.; Marro, T.; Paolucci, S.; Morelli, D. Stability and harmony of gait in children with cerebral palsy. Res. Dev. Disabil. 2012, 33, 129–135. [Google Scholar] [CrossRef] [PubMed]
  12. Lepage, C.; Noreau, L.; Bernard, P.-M. Association between characteristics of locomotion and accomplishment of life habits in children with cerebral palsy. Phys. Ther. 1998, 78, 458–469. [Google Scholar] [CrossRef] [Green Version]
  13. Sorsdahl, A.B.; Moe-Nilssen, R.; Strand, L.I. Test–retest reliability of spatial and temporal gait parameters in children with cerebral palsy as measured by an electronic walkway. Gait Posture 2008, 27, 43–50. [Google Scholar] [CrossRef] [Green Version]
  14. Booth, A.T.; Buizer, A.I.; Meyns, P.; Oude Lansink, I.L.; Steenbrink, F.; van der Krogt, M.M. The efficacy of functional gait training in children and young adults with cerebral palsy: A systematic review and meta-analysis. Dev. Med. Child Neurol. 2018, 60, 866–883. [Google Scholar] [CrossRef]
  15. Chakraborty, S.; Nandy, A.; Kesar, T.M. Gait deficits and dynamic stability in children and adolescents with cerebral palsy: A systematic review and meta-analysis. Clin. Biomech. 2020, 71, 11–23. [Google Scholar] [CrossRef] [PubMed]
  16. Mehdizadeh, S.; Dolatabadi, E.; Ng, K.-D.; Mansfield, A.; Flint, A.; Taati, B.; Iaboni, A. Vision-based assessment of gait features associated with falls in people with dementia. J. Gerontol. Ser. A 2020, 75, 1148–1153. [Google Scholar] [CrossRef] [PubMed]
  17. Ng, K.D.; Mehdizadeh, S.; Iaboni, A.; Mansfield, A.; Flint, A.; Taati, B. Measuring Gait Variables Using Computer Vision to Assess Mobility and Fall Risk in Older Adults With Dementia. IEEE J. Transl. Eng. Health Med. 2020, 8, 1–9. [Google Scholar] [CrossRef]
  18. Abd El-Kafy, E.M.; El-Basatiny, H.M.Y.M. Effect of postural balance training on gait parameters in children with cerebral palsy. Am. J. Phys. Med. Rehabil. 2014, 93, 938–947. [Google Scholar] [CrossRef]
  19. Elnahhas, A.M.; Elshennawy, S.; Aly, M.G. Effects of backward gait training on balance, gross motor function, and gait in children with cerebral palsy: A systematic review. Clin. Rehabil. 2019, 33, 3–12. [Google Scholar] [CrossRef]
  20. Moreau, N.G.; Bodkin, A.W.; Bjornson, K.; Hobbs, A.; Soileau, M.; Lahasky, K. Effectiveness of rehabilitation interventions to improve gait speed in children with cerebral palsy: Systematic review and meta-analysis. Phys. Ther. 2016, 96, 1938–1954. [Google Scholar] [CrossRef] [Green Version]
  21. Hof, A.; Gazendam, M.; Sinke, W. The condition for dynamic stability. J. Biomech. 2005, 38, 1–8. [Google Scholar] [CrossRef]
  22. Winter, D.A.; Patla, A.E.; Prince, F.; Ishac, M.; Gielo-Perczak, K. Stiffness Control of Balance in Quiet Standing. J. Neurophysiol. 1998, 80, 1211–1221. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  23. Young, P.M.M.; Wilken, J.M.; Dingwell, J.B. Dynamic margins of stability during human walking in destabilizing environments. J. Biomech. 2012, 45, 1053–1059. [Google Scholar] [CrossRef] [Green Version]
  24. Hak, L.; Houdijk, H.; Van Der Wurff, P.; Prins, M.R.; Mert, A.; Beek, P.J.; Van Dieën, J.H. Stepping strategies used by post-stroke individuals to maintain margins of stability during walking. Clin. Biomech. 2013, 28, 1041–1048. [Google Scholar] [CrossRef] [Green Version]
  25. Martelli, D.; Luo, L.; Kang, J.; Kang, U.J.; Fahn, S.; Agrawal, S.K. Adaptation of stability during perturbed walking in Parkinson’s disease. Sci. Rep. 2017, 7, 1–11. [Google Scholar] [CrossRef]
  26. van Vugt, Y.; Stinear, J.; Davies, T.C.; Zhang, Y. Postural stability during gait for adults with hereditary spastic paraparesis. J. Biomech. 2019, 88, 12–17. [Google Scholar] [CrossRef]
  27. Delabastita, T.; Desloovere, K.; Meyns, P. Restricted arm swing affects gait stability and increased walking speed alters trunk movements in children with cerebral palsy. Front. Hum. Neurosci. 2016, 10, 354. [Google Scholar] [CrossRef] [Green Version]
  28. Peebles, A.T.; Reinholdt, A.; Bruetsch, A.P.; Lynch, S.G.; Huisinga, J.M. Dynamic margin of stability during gait is altered in persons with multiple sclerosis. J. Biomech. 2016, 49, 3949–3955. [Google Scholar] [CrossRef] [Green Version]
  29. Ejupi, A.; Gschwind, Y.J.; Valenzuela, T.; Lord, S.R.; Delbaere, K. A kinect and inertial sensor-based system for the self-assessment of fall risk: A home-based study in older people. Hum.–Comput. Interact. 2016, 31, 261–293. [Google Scholar] [CrossRef]
  30. Zhong, R.; Rau, P.-L.P. Are cost-effective technologies feasible to measure gait in older adults? A systematic review of evidence-based literature. Arch. Gerontol. Geriatr. 2020, 87, 103970. [Google Scholar] [CrossRef]
  31. Bower, K.; Thilarajah, S.; Pua, Y.-H.; Williams, G.; Tan, D.; Mentiplay, B.; Denehy, L.; Clark, R. Dynamic balance and instrumented gait variables are independent predictors of falls following stroke. J. Neuroeng. Rehabil. 2019, 16, 3. [Google Scholar] [CrossRef] [PubMed]
  32. Springer, S.; Yogev Seligmann, G. Validity of the kinect for gait assessment: A focused review. Sensors 2016, 16, 194. [Google Scholar] [CrossRef] [PubMed]
  33. Clark, R.A.; Mentiplay, B.F.; Hough, E.; Pua, Y.H. Three-dimensional cameras and skeleton pose tracking for physical function assessment: A review of uses, validity, current developments and Kinect alternatives. Gait Posture 2019, 68, 193–200. [Google Scholar] [CrossRef] [PubMed]
  34. Stokes, E.K. 5—Validity, in Rehabilitation Outcome Measures; Stokes, E.K., Ed.; Churchill Livingstone: Edinburgh, UK, 2011; pp. 35–46. [Google Scholar]
  35. Mentiplay, B.F.; Perraton, L.G.; Bower, K.J.; Pua, Y.-H.; McGaw, R.; Heywood, S.; Clark, R.A. Gait assessment using the Microsoft Xbox One Kinect: Concurrent validity and inter-day reliability of spatiotemporal and kinematic variables. J. Biomech. 2015, 48, 2166–2170. [Google Scholar] [CrossRef] [PubMed]
  36. Latorre, J.; Llorens, R.; Colomer, C.; Alcañiz, M. Reliability and comparison of Kinect-based methods for estimating spatiotemporal gait parameters of healthy and post-stroke individuals. J. Biomech. 2018, 72, 268–273. [Google Scholar] [CrossRef] [PubMed]
  37. Barreira, C.C.; Forner-Cordero, A.; Grangeiro, P.M.; Moura, R.T. Kinect v2 based system for gait assessment of children with cerebral palsy in rehabilitation settings. J. Med. Eng. 2020, 44, 198–202. [Google Scholar] [CrossRef] [PubMed]
  38. Palisano, R.; Rosenbaum, P.; Walter, S.; Russell, D.; Wood, E.; Galuppi, B. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev. Med. Child Neurol. 1997, 39, 214–223. [Google Scholar] [CrossRef]
  39. Sutherland, D.H. The evolution of clinical gait analysis: Part II Kinematics. Gait Posture 2002, 16, 159–179. [Google Scholar] [CrossRef]
  40. Mackey, A.H.; Walt, S.E.; Lobb, G.A.; Stott, N.S. Reliability of upper and lower limb three-dimensional kinematics in children with hemiplegia. Gait Posture 2005, 22, 1–9. [Google Scholar] [CrossRef]
  41. Xu, X.; McGorry, R.W.; Chou, L.-S.; Lin, J.-H.; Chang, C.-C. Accuracy of the Microsoft Kinect™ for measuring gait parameters during treadmill walking. Gait Posture 2015, 42, 145–151. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  42. Zeni, J., Jr.; Richards, J.; Higginson, J. Two simple methods for determining gait events during treadmill and overground walking using kinematic data. Gait Posture 2008, 27, 710–714. [Google Scholar] [CrossRef] [Green Version]
  43. Bruening, D.A.; Ridge, S.T. Automated event detection algorithms in pathological gait. Gait Posture 2014, 39, 472–477. [Google Scholar] [CrossRef]
  44. Eltoukhy, M.; Oh, J.; Kuenze, C.; Signorile, J. Improved kinect-based spatiotemporal and kinematic treadmill gait assessment. Gait Posture 2017, 51, 77–83. [Google Scholar] [CrossRef]
  45. Summa, S.; Tartarisco, G.; Favetta, M.; Buzachis, A.; Romano, A.; Bernava, G.; Sancesario, A.; Vasco, G.; Pioggia, G.; Petrarca, M. Validation of low-cost system for gait assessment in children with ataxia. Comput. Methods Programs Biomed. 2020, 196, 105705. [Google Scholar] [CrossRef]
  46. Summa, S.; Tartarisco, G.; Favetta, M.; Buzachis, A.; Romano, A.; Bernava, G.; Vasco, G.; Pioggia, G.; Petrarca, M.; Castelli, E. Spatio-temporal parameters of ataxia gait dataset obtained with the Kinect. Data in Brief 2020, 32, 106307. [Google Scholar] [CrossRef]
  47. Whittle, M.W. Three-dimensional motion of the center of gravity of the body during walking. Hum. Mov. Sci. 1997, 16, 347–355. [Google Scholar] [CrossRef]
  48. Gibson, T.; Jeffery, R.; Bakheit, A. Comparison of three definitions of the mid-stance and mid-swing events of the gait cycle in children. Disabil. Rehabil. 2006, 28, 625–628. [Google Scholar] [CrossRef]
  49. Fleiss, J.L. Design and Analysis of Clinical Experiments; John Wiley & Sons: Hoboken, NJ, USA, 2011; Volume 73. [Google Scholar]
  50. Denegar, C.R.; Ball, D.W. Assessing reliability and precision of measurement: an introduction to intraclass correlation and standard error of measurement. J. Sport Rehabil. 1993, 2, 35–42. [Google Scholar] [CrossRef]
  51. Weir, J.P. Quantifying test-retest reliability using the intraclass correlation coefficient and the SEM. Strength. Cond. J. 2005, 19, 231–240. [Google Scholar]
  52. Bland, J.M.; Altman, D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986, 327, 307–310. [Google Scholar] [CrossRef]
  53. Van Bloemendaal, M.; Beelen, A.; Kleissen, R.F.; Geurts, A.C.; Nollet, F.; Bus, S.A. Concurrent validity and reliability of a low-cost gait analysis system for assessment of spatiotemporal gait parameters. J. Rehabil. Med. 2019, 51, 456–463. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  54. Dolatabadi, E.; Taati, B.; Mihailidis, A. Concurrent validity of the Microsoft Kinect for Windows v2 for measuring spatiotemporal gait parameters. Med. Eng. Phys. 2016, 38, 952–958. [Google Scholar] [CrossRef]
  55. Clark, R.A.; Vernon, S.; Mentiplay, B.F.; Miller, K.J.; McGinley, J.L.; Pua, Y.H.; Paterson, K.; Bower, K.J. Instrumenting gait assessment using the Kinect in people living with stroke: reliability and association with balance tests. J. Neuroeng. Rehabil. 2015, 12, 15. [Google Scholar] [CrossRef] [PubMed]
  56. Otte, K.; Kayser, B.; Mansow-Model, S.; Verrel, J.; Paul, F.; Brandt, A.U.; Schmitz-Hübsch, T. Accuracy and reliability of the kinect version 2 for clinical measurement of motor function. PLoS ONE 2016, 11, e0166532. [Google Scholar] [CrossRef]
  57. Grobelny, A.; Behrens, J.R.; Mertens, S.; Otte, K.; Mansow-Model, S.; Kruger, T.; Gusho, E.; Bellmann-Strobl, J.; Paul, F.; Brandt, A.U.; et al. Maximum walking speed in multiple sclerosis assessed with visual perceptive computing. PLoS ONE 2017, 12, e0189281. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  58. Latorre, J.; Colomer, C.; Alcañiz, M.; Llorens, R. Gait analysis with the Kinect v2: normative study with healthy individuals and comprehensive study of its sensitivity, validity, and reliability in individuals with stroke. J. Neuroeng. Rehabil. 2019, 16, 97. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  59. Steinert, A.; Sattler, I.; Otte, K.; Röhling, H.; Mansow-Model, S.; Müller-Werdan, U. Using new camera-based technologies for gait analysis in older adults in comparison to the established GAITRite system. Sensors 2020, 20, 125. [Google Scholar] [CrossRef] [Green Version]
  60. Wren, T.A.; Rethlefsen, S.; Kay, R.M. Prevalence of specific gait abnormalities in children with cerebral palsy: influence of cerebral palsy subtype, age, and previous surgery. J. Pediatr. Orthop. 2005, 25, 79–83. [Google Scholar] [PubMed]
  61. Auvinet, E.; Multon, F.; Manning, V.; Meunier, J.; Cobb, J. Validity and sensitivity of the longitudinal asymmetry index to detect gait asymmetry using Microsoft Kinect data. Gait Posture 2017, 51, 162–168. [Google Scholar] [CrossRef]
  62. Auvinet, E.; Multon, F.; Aubin, C.-E.; Meunier, J.; Raison, M. Detection of gait cycles in treadmill walking using a Kinect. Gait Posture 2015, 41, 722–725. [Google Scholar] [CrossRef]
  63. Motiian, S.; Pergami, P.; Guffey, K.; Mancinelli, C.A.; Doretto, G. Automated extraction and validation of children’s gait parameters with the Kinect. Biomed. Eng. Online 2015, 14, 112. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  64. Oh, J.; Kuenze, C.; Jacopetti, M.; Signorile, J.F.; Eltoukhy, M. Validity of the microsoft kinect™ in assessing spatiotemporal and lower extremity kinematics during stair ascent and descent in healthy young individuals. Med. Eng. Phys. 2018, 60, 70–76. [Google Scholar] [CrossRef] [PubMed]
  65. de Jong, L.; van Dijsseldonk, R.; Keijsers, N.; Groen, B. Test-retest reliability of stability outcome measures during treadmill walking in patients with balance problems and healthy controls. Gait Posture 2020, 76, 92–97. [Google Scholar] [CrossRef]
  66. Redekop, S.; Andrysek, J.; Wright, V. Single-session reliability of discrete gait parameters in ambulatory children with cerebral palsy based on GMFCS level. Gait Posture 2008, 28, 627–633. [Google Scholar] [CrossRef] [PubMed]
  67. Ghai, S.; Ghai, I. Virtual reality enhances gait in cerebral palsy: A training dose-response meta-analysis. Front. Neurol. 2019, 10, 236. [Google Scholar] [CrossRef]
  68. Dumas, R.; Cheze, L.; Verriest, J.-P. Adjustments to McConville et al. and Young et al. body segment inertial parameters. J. Biomech. 2007, 40, 543–553. [Google Scholar] [CrossRef]
  69. Devetak, G.F.; Bohrer, R.C.D.; Rodacki, A.L.F.; Manffra, E.F. Center of mass in analysis of dynamic stability during gait following stroke: A systematic review. Gait Posture 2019, 72, 154–166. [Google Scholar] [CrossRef]
  70. González, A.; Hayashibe, M.; Bonnet, V.; Fraisse, P. Whole body center of mass estimation with portable sensors: Using the statically equivalent serial chain and a kinect. Sensors 2014, 14, 16955–16971. [Google Scholar] [CrossRef] [Green Version]
  71. Skals, S.; Rasmussen, K.P.; Bendtsen, K.M.; Yang, J.; Andersen, M.S. A musculoskeletal model driven by dual Microsoft Kinect Sensor data. Multibody Syst. Dyn. 2017, 41, 297–316. [Google Scholar] [CrossRef] [Green Version]
  72. Jansi, R.; Amutha, R. Detection of fall for the elderly in an indoor environment using a tri-axial accelerometer and Kinect depth data. Multidimens. Syst. Signal Process. 2020, 31, 1207–1225. [Google Scholar] [CrossRef]
  73. Rodrigues, T.B.; Salgado, D.P.; Catháin, C.Ó.; O’Connor, N.; Murray, N. Human gait assessment using a 3D marker-less multimodal motion capture system. Multimed. Tools Appl. 2020, 79, 2629–2651. [Google Scholar] [CrossRef]
  74. Albert, J.A.; Owolabi, V.; Gebel, A.; Brahms, C.M.; Granacher, U.; Arnrich, B. Evaluation of the Pose Tracking Performance of the Azure Kinect and Kinect v2 for Gait Analysis in Comparison with a Gold Standard: A Pilot Study. Sensors 2020, 20, 5104. [Google Scholar] [CrossRef] [PubMed]
Sensors 21 02104 g001 550
Figure 1. Lower limb reflective marker locations of a modified Cleveland Clinic marker set (a) and joint landmarks recorded by the Kinect V2 sensor (b).
Figure 1. Lower limb reflective marker locations of a modified Cleveland Clinic marker set (a) and joint landmarks recorded by the Kinect V2 sensor (b).
Sensors 21 02104 g001
Sensors 21 02104 g002 550
Figure 2. Illustration of the calculation of margin of stability (MOS).
Figure 2. Illustration of the calculation of margin of stability (MOS).
Sensors 21 02104 g002
Sensors 21 02104 g003 550
Figure 3. Bland–Altman plots indicate the agreement between MOS variables for ten children with cerebral palsy (CP) calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.
Figure 3. Bland–Altman plots indicate the agreement between MOS variables for ten children with cerebral palsy (CP) calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.
Sensors 21 02104 g003
Sensors 21 02104 g004 550
Figure 4. Bland–Altman plots indicate the agreement between gait spatiotemporal variables for ten children with CP calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.
Figure 4. Bland–Altman plots indicate the agreement between gait spatiotemporal variables for ten children with CP calculated using the Kinect V2 and Motion Analysis. The red solid line represents the reference line at the mean, and the two black dashed lines represent the upper and lower limit of agreement.
Sensors 21 02104 g004
Table
Table 1. Definitions of spatiotemporal gait parameters in the Kinect V2 and Motion Analysis system [35,45,46].
Table 1. Definitions of spatiotemporal gait parameters in the Kinect V2 and Motion Analysis system [35,45,46].
Motion AnalysisKinect V2
Step Length (m)Distance between the heel markers at the left and right foot strikeDistance between the ankles at the left and right foot strike
Stride Length (m)Distance between RHEEs at the two consecutive right foot strikeDistance between the “ankle right” markers at the two consecutive right foot strike
Step Width (m)Orthogonal distance from the LHEE to the vector formed by RHEEs in two consecutive foot strikeOrthogonal distance from the left ankles to the vector formed by the right ankles in two consecutive foot strike
Gait Speed (m/s)Mean resultant velocity of the COM during the gait cycleMean resultant velocity of the “spine base” marker during the gait cycle
Step Time (s)The time between the left and right foot strikeAs per the Motion Analysis
Stride Time (s)The time between two consecutive right foot strikeAs per the Motion Analysis
Table
Table 2. List of calculation formulas of spatiotemporal gait parameters in the Kinect V2 and Motion Analysis system [35,45,46]. 3D coordinates of RHEE in Motion Analysis or “right ankle” in Kinect V2 at the first and second right foot strike were represented by x 1 , y 1 , z 1 and x 3 , y 3 , z 3 , respectively. 3D coordinates of LHEE in Motion Analysis or “left ankle” in Kinect V2 at the left foot strike were represented by x 2 , y 2 , z 2 . Average positions of the sacrum, LASI and RASI or “spine base” in the Kinect V2 were represented by x C O M , y C O M , z C O M . t 1 , t 2 , and t 3 stood for time frames of first right foot strike, the left foot strike, and the second right foot strike. ρ stood for the sampling frequency of Motion Analysis or Kinect V2.
Table 2. List of calculation formulas of spatiotemporal gait parameters in the Kinect V2 and Motion Analysis system [35,45,46]. 3D coordinates of RHEE in Motion Analysis or “right ankle” in Kinect V2 at the first and second right foot strike were represented by x 1 , y 1 , z 1 and x 3 , y 3 , z 3 , respectively. 3D coordinates of LHEE in Motion Analysis or “left ankle” in Kinect V2 at the left foot strike were represented by x 2 , y 2 , z 2 . Average positions of the sacrum, LASI and RASI or “spine base” in the Kinect V2 were represented by x C O M , y C O M , z C O M . t 1 , t 2 , and t 3 stood for time frames of first right foot strike, the left foot strike, and the second right foot strike. ρ stood for the sampling frequency of Motion Analysis or Kinect V2.
Motion Analysis
Step Length (m) x 2 − x 1 2 + y 2 − y 1 2 + z 2 − z 1 2 (1)
Stride Length (m) x 3 − x 1 2 + y 3 − y 1 2 + z 3 − z 1 2 (2)
Step Width (m) x 2 − x 1 , y 2 − y 1 , z 2 − z 1 × x 3 − x 1 , y 3 − y 1 , z 3 − z 1 x 3 − x 1 , y 3 − y 1 , z 3 − z 1 (3)
Gait Speed (m/s) ρ 2 ( t 3 − t 1 ) ∑ i = 1 n x i + 1 , C O M − x i − 1 , C O M 2 + y i + 1 , C O M − y i − 1 , C O M 2 + z i + 1 , C O M − z i − 1 , C O M 2 (4)
Step Time (s) t 2 − t 1 ρ (5)
Stride Time (s) t 3 − t 1 ρ (6)
Table
Table 3. Mean ± 1SD (n = 10) values for each MOS and spatiotemporal gait parameters collected by Motion Analysis and the Kinect on Day 1, intraclass correlation coefficient (ICC2,k) with 95% confidence intervals for agreement, SEM, and relative errors (in percentage) are provided. FS: foot strike; min: minimum; L: length of the pendulum.
Table 3. Mean ± 1SD (n = 10) values for each MOS and spatiotemporal gait parameters collected by Motion Analysis and the Kinect on Day 1, intraclass correlation coefficient (ICC2,k) with 95% confidence intervals for agreement, SEM, and relative errors (in percentage) are provided. FS: foot strike; min: minimum; L: length of the pendulum.
Motion AnalysisKinectICC2,kSEMRelative Errors (%)
MOS-ML-min (m)0.05 ± 0.020.04 ± 0.020.81 (0.20, 0.95)0.00432.04 ± 23.06
MOS-ML-mid stance (m)0.09 ± 0.030.06 ± 0.030.68 (−0.25, 0.93)0.0137.06 ± 19.61
MOS-ML-FS (m)0.08 ± 0.020.09 ± 0.010.42 (−1.78, 0.86)0.00422.06 ± 27.15
MOS-AP-min (m)−0.26 ± 0.13−0.26 ± 0.130.99 (0.96, 1.00)0.039.83 ± 7.51
MOS-AP-mid stance (m)−0.06 ± 0.05−0.14 ± 0.070.66 (−0.21, 0.92)0.06112.82 ± 50.42
MOS-AP-FS (m)0.02 ± 0.03−0.03 ± 0.050.51 (−0.29, 0.86)0.03186.40 ± 163.49
Step Length (m)0.33 ± 0.080.34 ± 0.080.99 (0.94, 0.99)0.014.47 ± 4.62
Stride Length (m)0.55 ± 0.200.55 ± 0.200.99 (0.99, 0.99)0.022.49 ± 1.75
Step Width (m)0.16 ± 0.040.18 ± 0.030.83 (0.17, 0.96)0.0116.81 ± 11.24
Gait Speed (m/s)0.60 ± 0.220.60 ± 0.230.99 (0.99, 0.99)0.032.69 ± 2.12
Step Time (s)0.47 ± 0.080.47 ± 0.070.98 (0.94, 0.99)0.013.23 ± 2.21
Stride Time (s)0.97 ± 0.130.96 ± 0.130.99 (0.98, 0.99)0.021.75 ± 0.72
L (m)0.57 ± 0.060.42 ± 0.080.45 (−0.07, 0.85)0.1425.97 ± 6.84
Table
Table 4. Results of Bland–Altman analysis of agreement between gait parameters calculated from the Kinect and Motion Analysis.
Table 4. Results of Bland–Altman analysis of agreement between gait parameters calculated from the Kinect and Motion Analysis.
ParameterMean DifferenceLoALower LoAUpper LoA
MOS-ML-min (m)0.010.03−0.020.04
MOS-ML-mid stance (m)0.030.04−0.010.07
MOS-ML-FS (m)−0.0020.04−0.040.04
MOS-AP-min (m)−0.0020.05−0.050.05
MOS-AP-mid stance (m)0.070.08−0.0020.15
MOS-AP-FS (m)0.050.07−0.020.12
Step Length (m)−0.010.03−0.040.03
Stride Length (m)−0.010.03−0.030.02
Step Width (m)−0.020.04−0.060.02
Gait Speed (m/s)−0.0040.03−0.040.03
Step Time (s)0.010.04−0.030.04
Stride Time (s)0.0040.04−0.030.04
L (m)0.150.080.070.22
Table
Table 5. Mean ± 1SD (n = 10) values for each MOS and spatiotemporal gait parameters collected by the Kinect V2 on Day 1 and 2, ICC2,k with 95% confidence intervals for agreement and SEM is provided. FS: foot strike; min: minimum; L: length of the pendulum.
Table 5. Mean ± 1SD (n = 10) values for each MOS and spatiotemporal gait parameters collected by the Kinect V2 on Day 1 and 2, ICC2,k with 95% confidence intervals for agreement and SEM is provided. FS: foot strike; min: minimum; L: length of the pendulum.
Day 1Day 2ICC2,kSEM
MOS-ML-min (m)0.03 ± 0.030.02 ± 0.030.63 (−0.22, 0.90)0.01
MOS-ML-mid stance (m)0.05 ± 0.040.03 ± 0.020.28 (−1.47, 0.81)0.01
MOS-ML-FS (m)0.08 ± 0.030.08 ± 0.030.56 (−1.01, 0.89)0.01
MOS-AP-min (m)−0.22 ± 0.11−0.31 ± 0.100.64 (−0.24, 0.91)0.15
MOS-AP-mid stance (m)−0.17 ± 0.06−0.14 ± 0.070.69 (−0.06, 0.92)0.05
MOS-AP-FS (m)−0.03 ± 0.070.01 ± 0.050.52 (−0.54, 0.87)0.05
Step Length (m)0.40 ± 0.080.40 ± 0.100.82 (0.23, 0.96)0.06
Stride Length (m)0.71 ± 0.190.70 ± 0.170.83 (0.27, 0.96)0.24
Step Width (m)0.17 ± 0.040.16 ± 0.040.88 (0.56, 0.97)0.01
Gait Speed (m/s)0.76 ± 0.230.76 ± 0.250.78 (0.06, 0.95)0.49
Step Time (s)0.49 ± 0.090.48 ± 0.060.85 (0.41, 0.96)0.04
Stride Time (s)1.00 ± 0.150.96 ± 0.130.82 (0.35, 0.96)0.15
L (m)0.44 ± 0.050.43 ± 0.040.92 (0.69, 0.98)0.01
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Ma, Y.; Mithraratne, K.; Wilson, N.; Zhang, Y.; Wang, X. Kinect V2-Based Gait Analysis for Children with Cerebral Palsy: Validity and Reliability of Spatial Margin of Stability and Spatiotemporal Variables. Sensors 2021, 21, 2104. https://doi.org/10.3390/s21062104

AMA Style

Ma Y, Mithraratne K, Wilson N, Zhang Y, Wang X. Kinect V2-Based Gait Analysis for Children with Cerebral Palsy: Validity and Reliability of Spatial Margin of Stability and Spatiotemporal Variables. Sensors. 2021; 21(6):2104. https://doi.org/10.3390/s21062104

Chicago/Turabian Style

Ma, Yunru, Kumar Mithraratne, Nichola Wilson, Yanxin Zhang, and Xiangbin Wang. 2021. "Kinect V2-Based Gait Analysis for Children with Cerebral Palsy: Validity and Reliability of Spatial Margin of Stability and Spatiotemporal Variables" Sensors 21, no. 6: 2104. https://doi.org/10.3390/s21062104

APA Style

Ma, Y., Mithraratne, K., Wilson, N., Zhang, Y., & Wang, X. (2021). Kinect V2-Based Gait Analysis for Children with Cerebral Palsy: Validity and Reliability of Spatial Margin of Stability and Spatiotemporal Variables. Sensors, 21(6), 2104. https://doi.org/10.3390/s21062104

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop