Received: 27 May 2018 Revised: 20 May 2019 Accepted: 4 July 2019

DOI: 10.1002/pri.1803

RESEARCH ARTICLE

Concurrent validity and intratester reliability of the video-

based system for measuring gait poststroke

Nilar Aung1,2 | Sunee Bovonsunthonchai1 | Vimonwan Hiengkaew1 |

1 1
Jarugool Tretriluxana | Rommanee Rojasavastera | Anuchai Pheung-Phrarattanatrai3

1
Faculty of Physical Therapy, Mahidol
University, Nakhon Pathom, Thailand Abstract
2
Department of Physiotherapy, University of Background and purpose: Spatio-temporal parameters are commonly used in gait
Medical Technology, Yangon, Myanmar
assessment. Advanced tools provide valid and reliable data, considered very effective
3
Physical Therapy Center, Faculty of Physical
Therapy, Mahidol University, Bangkok, for physiotherapy intervention. However, these tools may be limited in clinical usage
Thailand caused by complicated applicability, inaccessibility, and high cost. Therefore, a video-

Correspondence based system is an alternative choice that is easy and affordable for the clinical set-
Sunee Bovonsunthonchai, Faculty of Physical ting. The purpose of the study was to evaluate the concurrent validity of the video-
Therapy, Mahidol University, 999
Phuttamonton Sai 4 Rd., Salaya, based system against the validated instrumented gait system (Force Distribution
Phuttamonthon, Nakhon Pathom, 73170, Measurement [FDM]) on the spatio-temporal gait parameters in individuals with
Thailand.
Email: sunee.bov@mahidol.ac.th stroke. In addition, the intratester reliability of a novice tester was determined.
Methods: Twenty individuals with stroke participated in the study. Gait was captured
Funding information
the Norway Scholarship (Mahidol-Norway by the video-based and FDM systems simultaneously to measure the degree of con-
Capacity Building Initiative for ASEAN). current validity. Parameters composed of the affected and unaffected step lengths
(cm) and step time (s), stride length (cm), gait velocity (m/s), and cadence (steps/min).
Pearson correlation coefficient, paired t test, and intraclass correlation coefficient
(ICC) were used to determine the concurrent validity, the difference of the data, and
intratester reliability.
Results: All spatio-temporal gait parameters showed excellent degrees of correlation
(rp = .94 to.99, p <.001) between the video-based and FDM systems. No significant
difference in all parameters was found between the two systems. Excellent
intratester reliability (ICC3,1 = 0.91 to 0.99, p < .001) of all gait parameters were
found in a novice tester.
Conclusion: The video-based system was valid and reliable for a novice tester to
measure the spatio-temporal gait parameters in individuals with stroke.

KEYWORDS
gait analysis, reproducibility of results, stroke, video recording

1 | I N T RO D UC T I O N concerns for physiotherapy practice. From the 1,836 physiotherapists

who answered the questionnaire, 66.4% replied that they needed
From the previous survey (Toro, Nester, & Farren, 2003) about gait training in gait assessment and 91.8% replied that they needed the
assessment of therapists in the National Health Service in the United tools that can be used easily and quickly, preserving reliability and
Kingdom, management of the abnormal gait is one of the major validity (Toro et al., 2003). Traditionally, simple methods such as the

Physiother Res Int. 2019;e1803. wileyonlinelibrary.com/journal/pri © 2019 John Wiley & Sons, Ltd. 1 of 7
https://doi.org/10.1002/pri.1803
2 of 7 AUNG ET AL.

paper-and-pencil method (Sekiya, Nagasaki, Ito, & Furuna, 1997), second purpose was to determine the intratester reliability of a novice
visual observation (Krebs, Edelstein, & Fishman, 1985), and stopwatch tester.
(Wall & Scarbrough, 1997) were usually used to evaluate gait abnor-
malities. However, there are limitations of these methods such as
2 | METHODS
labour-intensive, time-consuming, and restricted information on valid-
ity and reliability. Establishment of the sophisticated equipment such
A cross-sectional design with a convenience sampling technique was
as force plate, 3D motion analysis, body-mounted accelerometer, and
used in the study. Individuals with stroke from the Physical Therapy
instrumented walkway mat have been raised (Bilney, Morris, & Web-
Center, Faculty of Physical Therapy, Mahidol University, participated
ster, 2003; Henriksen, Lund, Moe-Nilssen, Bliddal, & Danneskiod-
in the study. Prior to participating in the study, participants signed
Samsoe, 2004; Moore et al., 2017; Tanikawa et al., 2016). However,
informed consents approved by the institutional research ethical com-
utilization of this equipment may be limited because of low accessibil-
mittee (COA: MU-CIRB 2017/178.1010).
ity, high cost, sensor fragility, and operating complexity. To provide
the confidence in a simple tool used to detect gait abnormality, testing
of validity and reliability is necessary. 2.1 | Participants
A Force Distribution Measurement (FDM) platform, a type of
Selection criteria of the participants included ischemic or hemorrhagic
the instrumented walkway system, was developed to measure gait
stroke, first stroke with unilateral involvement, and ability to walk
in clinics. It was proven to be a valid and reliable method to mea-
without using orthotic devices such as ankle–foot orthotic or foot slap
sure gait variables (Giacomozzi, 2010) and useful for rehabilitation
for at least 8 m. The participants who had (a) structural leg length dis-
in several conditions (Bovonsunthonchai et al., 2018; Kalron, Dvir,
crepancy of more than 2 cm, (b) opened wounds and/or infection at
Givon, Baransi, & Achiron, 2014; Wollesen, Voelcker-Rehage, Wil-
their feet, (c) musculoskeletal or cardiopulmonary conditions that may
ler, Zech, & Mattes, 2015). After collecting data, FDM was able to
interfere with walking ability or injury during testing were excluded
display the results without being labour-intensive. However, it may
from the study. Due to the limitations of measurement tool and
not be appropriate to use it in the clinical setting because it costs
method of event identification, the FDM investigated gait parameters
high, it is not portable, and the pressure sensors may be damaged
on the basis of a pressure sensor, and the video-based system
by gait aids or hard treads.
required the tester to identify the initial contact and foot-off events;
A video-based assessment is an observational method for motion
the participants who had severe foot drop or foot drag over the whole
analysis that allows therapists to analyse gait repeatedly with a slow-
gait cycle were excluded from the study.
motion shot and an ability to stop for a static pose. A video is very
helpful because it is simple, portable, and affordable to determine gait
characteristics in both qualitative and quantitative data. It has been 2.2 | Instrumentation and laboratory configuration
tested for the validity and reliability in several tasks on the basis of
The FDM system (Zebris, Germany) composed of an electronic walk-
joint angles (Munro, Herrington, & Carolan, 2012; Norris & Olson,
way with the size of 307 × 60.5 × 2.1 cm (length × width × height)
2011). The study of its concurrent validity with the 3D motion analy-
connected to a personal computer was used as a gold standard
sis system showed a strong correlation (R2 = .99) for determining kine-
measurement. For the video-based system, a video camera (Sanyo
matics gait data in healthy subjects (Ugbolue et al., 2013). However,
Xacti VPC-GH1, Vietnam) was used and placed perpendicular to the
the study of Cutlip et al. showed differences in step length (p = .003)
and stride velocity (p = .0002) between the video-based system and participants. The walkway with a total length of 7 m was set, divided

GAITRite measurement (Cutlip, Mancinelli, Huber, & DiPasquale, into 3 m of the electronic walkway in the middle and 2 m of a com-

2000). Based on the study of concurrent validity between these two mon walkway before and after the electronic walkway.

tools, a low correlation was found in the right step length (intraclass Two 55-cm paper tapes were used as the calibrated rulers, adher-

correlation coefficient [ICC] = 0.44), whereas good to excellent corre- ing on the middle part of the electronic walkway parallel to the left

lations were found in the left step length (ICC = 0.85), step time (ICC and right footpaths to reduce the perspective error. To reduce the

= 0.97 for right; ICC = 0.95 for left), gait speed (ICC = 0.95), and parallax error, a video camera was mounted on a tripod stand, placed
cadence (ICC = 0.96) (McDonough, Batavia, Chen, Kwon, & Ziai, at the lateral and perpendicular to the participants and 3 m away from
2001). Different findings between studies may result from several fac- the walkway. With this distance, the sagittal gait motion can be cap-
tors such as the tools used, participant's characteristics, data collec- tured for two to three gait cycles, depending on the heights of the
tion, tracking, and analysis processes. participants. The height of the tripod stand was adjusted to capture
Until present, the study of the validity of a video-based system for the lower half of the participant. It was set at the lateral aspect of the
gait in individuals with stroke is very limited. Therefore, the purpose knee joint (approximately 53 cm). Moreover, the camera zoom was
of this study was to examine the concurrent validity of the video- adjusted to allow the two third of the walkway to be within the cam-
based system with a validated instrumented walkway system (FDM) era's field of view. The position of the video camera setting was
for the spatio-temporal gait parameters in individuals with stroke. The unaltered for the whole process of data collection. In addition, a
AUNG ET AL. 3 of 7

calibration process of the FDM system was performed before data velocity, and cadence were automatically calculated on the basis of
were initially collected. the pressure sensors of the platform and exported with the WinFDM
software version 1.1.2. In this study, data were digitized from the
2.3 | Validation of the video-based system with the same steps with the video-based system.
FDM system For the video-based system, recorded videos were opened by the
Kinovea software version 0.8.15 to calculate step lengths and step
Validation of the spatio-temporal gait data from the video-based sys-
times. The data of one gait cycle in the middle part of the walkway
tem was investigated with the data extracting from FDM. Three con-
were selected. To measure step length, initial contacts of the right and
secutive trials for preferred and fast speeds were recorded. The
left feet were marked by using forward and backward motions' func-
averaged values from three consecutive trials were calculated.
tion. The exact values of step length were calculated on the basis of
the calibrated rulers (reference known distance) that were placed on
2.4 | Intratester reliability of the video-based system the walkway. It was calculated by the relation of the distance (cm) of
Prior to collecting the data, a novice tester who had no experience the calibrated ruler with the distance (pixels) of the image. All data cal-
with the tool was trained on equipment setting, data tracking, culation from the video-based system was performed by one tester.
processing, and reporting with the expert for 2 weeks. The novice tes- For measurement of the right and left step times, it was calculated as
ter was a physiotherapist who had over 5 years of experience in the difference of the time from initial contact of one foot to the con-
treating patients with neurological conditions. tralateral foot contact. The rest of the parameters (stride length,
To investigate the intratester reliability, gait data were captured velocity, and cadence) were calculated by using the derived formula:
for two sessions on the same day because gait behaviours in individ-
uals with stroke can change easily over days. To avoid the learning Stride length = left step length + right step length,
effect from the first session of testing, a 30-min rest period was
allowed between testing sessions. For each testing session, three con- Stride time = left step time + right step time,
secutive trials of gait were measured for preferred and fast speeds. A
5-min break was provided between conditions of speed to prevent Velocity = stride length=stride time,
fatigue. The averaged values from three consecutive trials were
calculated. Cadence = speed*2*60=stride length:

2.5 | Procedures for data collection For the statistical analysis, IBM SPSS statistic version 19 was used
to calculate the data. The p value <.05 was considered statistically sig-
Age, gender, height, weight, and stroke characteristics were recorded
nificant in all analyses. Descriptive statistic was used to report the
for each participant prior to collecting the data. The two assessment
participant characteristics. Normal distribution of the data was
systems recorded gait simultaneously. To capture preferred and fast
assessed by using the Kolmogorov–Smirnov goodness-of-fit test. The
speeds of walking, the data were collected with 100 Hz for the FDM
Pearson correlation statistics were used to determine the concurrent
and 60 Hz for the video-based camera.
validity of the data between the video-based and FDM systems. The
To reduce the variability of gait, all participants were instructed to
paired t test was used to determine the difference of data from two
start walking with the left leg and to practise walking until they were
assessment systems. ICCs with the two-way mixed effects, consis-
familiar with the environment and equipment. During practice, the
tency, single-measurement model (ICC3,1) and 95% confidence inter-
exact initiation point was adjusted in–out until a gait cycle presented
val were used to examine the intratester reliability and confidence
in the middle part of the 3-m electronic walkway. This was set easily
level of a novice tester for the video-based system data. In addition,
by attaching another marking line at the middle part of the 3-m walk-
the standard error of measurement (SEM) was estimated from the for-
way. Once the initiation point was clearly determined, all three trials pffiffiffiffiffiffiffi
mula SEM = σ 2e , where σ 2e is the error variance and equals the mean
of data collection were started at the same location. All participants
square error term from an ANOVA (Stratford & Goldsmith, 1997).
were asked to follow the same instructions “walk with face forward
towards the end of the platform at your preferred speed (or fast
speed), start walking with your left leg first, and start walking when I 2.7 | Data interpretation
say start and end walking when I say stop.” Data from one gait cycle
According to the guideline for the validity of measurement (Portney &
from the middle part of the walkway were selected to avoid accelera-
Watkins, 2015), interpretations of Pearson correlation coefficients
tion and deceleration effects.
were as follows: r < .25 (little or no relationship), r > .25 to r < .50 (fair
relationship), r > .50 to r < .75 (moderate to good relationship), and r >
2.6 | Data calculation and statistical analyses
.75 good to excellent relationship.
For the FDM system, spatio-temporal gait parameters including the Criteria for the reliability interpretation followed the guideline of
affected and unaffected step length and step time, stride length, gait Landis and Koch (1977) as ICC < 0.40 (poor to fair), ICC of 0.41 to
4 of 7 AUNG ET AL.

0.60 (moderate), ICC of 0.61 to 0.80 (excellent), and ICC of 0.81 to Pearson correlation ranged .94 to .99 (excellent) for preferred speed
1.00 (almost perfect). and ranged .99 to .99 (excellent) for fast speed of all gait parameters.
There were significant correlations of all gait parameters in both pre-
2.8 | Sample size calculation ferred and fast gait speeds.

The sample size of this study was estimated from our own pilot data
3.3 | Comparisons of the spatio-temporal gait
in individuals with stroke (n = 10) by using the G*Power 3.1 software
parameters between the video-based and FDM
version 3.1.9.2. On the basis of the step time of the right leg, the cor-
systems
relation coefficient was found at .90, and alpha level of .05 and power
of .90 were set in the calculation. The required number of sample was Table 3 presents the comparisons of gait parameters between the
8, so 20 participants in this study was sufficient. video-based and FDM systems. Means and standard deviations of all
parameters showed similarity for both preferred and fast speeds from
3 | RESULTS two assessment systems. No significant difference (p>.05) of the data
between these two systems was found.
3.1 | Participant characteristics
3.4 | Intratester reliability of the video-based system
Demographics of the participants are presented in Table 1. Twenty
individuals with stroke participated in the study; 13 had an ischemic Table 4 presents the intratester reliability of a novice tester for the
stroke, and seven had a hemorrhagic stroke. They were 12 males and video-based system. A novice tester showed excellent intratester reli-
eight females. Averaged age, weight, and height were 64.15 ± 10.55 ability in measuring all gait parameters using the video-based system
years, 64.32 ± 12.78 kg, 163.00 ± 10.00 cm, respectively. Mean post- in individuals with stroke during walking in both preferred and fast
stroke duration was 25.79 ± 22.70 months. The affected side, domi- gait speeds. The ICC3,1 ranged from 0.91 to 0.99 (almost perfect) for
nant side, physical activity level, and walking aids used in daily life preferred speed and from 0.97 to 0.99 (almost perfect) for fast speed.
were reported with the number of participants. Physical activity level To express the extent of the expected error of the rater, the SEM was
was collected from the history taking and divided into the active and described and ranged from 0.06 to 7.60 m/s for preferred speed and
the sedentary following the recommended physical activity level by from 0.04 to 6.50 m/s for fast speed.
the World Health Organization. The individuals who reported to prac-
tise walking, jogging, or other exercises for 3–4 hr/week were 4 | DISCUSSION
recorded as the physically active, and those who reported no or little
physical activity such as sitting, reclining, or lying down, which has a The aim of the present study was to determine concurrent validity
very low energy expenditure, were reported as the sedentary (World and intra-tester reliability of the video-based system on spatio-
Health Organization, 2010). temporal parameters in individuals with stroke. The spatio-temporal
gait parameters are the most commonly assessed and presented with
3.2 | Concurrent validity of the video-based system asymmetry (Balaban & Tok, 2014) in individuals with stroke. In the
with the FDM system clinical setting, a quantitative assessment is recognized as a useful tool
for gait disturbances providing functional diagnosis and treatment
Table 2 presents the correlation coefficients of gait parameters
plan and monitoring the disease progression. Thus, appropriateness of
between two assessment systems. Both data collection sessions,
gait analysis protocol, validity, and reliability obtained are required
TABLE 1 Demographics of the participants (n = 20) (Baker, Esquenazi, Benedetti, & Desloovere, 2016).
Characteristics Values (Mean ± SD or n) On the basis of the data from FDM, the present study found that

Gender (n) Male: 12, female: 8 the video-based system was valid to determine step length (r= .99),
step time (r = .99), stride length (r = .99), velocity (r = .94 to .99), and
Age (years) 64.15 ± 10.55
cadence (r = .97 to .99) in individuals with stroke during walking at the
Weight (kg) 64.32 ± 12.78
preferred and fast speeds. In addition, when comparing gait data from
Height (cm) 163.00 ± 10.00
these two systems, no significant difference (p > .05) in all gait param-
Stroke etiology (n) Ischemic: 13, hemorrhagic: 7
eters was found. These results were similar to the previous study per-
Stroke onset (months) 25.79 ± 22.70
formed in healthy subjects for the aspect of concurrent validity of the
Affected side (n) Left:11, right: 9
video-based system against the instrumented walkway (Cutlip et al.,
Dominant side (n) Left: 3, right:17
2000). The study of Cutlip et al. in 2000 found high Pearson correla-
Physical activity level (n) Active: 8, sedentary: 12
tions for step length (r = .94), step period (r = .97), stride velocity (r =
Walking aids (n) Using: 5 (single-point cane: 2, .99), stance duration (r = .99), and swing duration (r = .95) between
walker: 3), not using: 15
the two systems. In addition, comparisons of the two systems showed
Functional Ambulatory Category Level 3: 1, Level 4: 6, Level 5: 13
similar results of step length (63.40 cm for the pressure mat system
AUNG ET AL. 5 of 7

TABLE 2 Correlation coefficients of gait parameters between two assessment systems (n = 20)

Preferred speed Fast speed

1st session 2nd session 1st session 2nd session

Gait parameter rp p value rp p value rp p value rp p value

Step length (affected) .99 <.001 .99 <.001 .99 <.001 .99 <.001
Step length (unaffected) .99 <.001 .99 <.001 .99 <.001 .99 <.001
Step time (affected) .99 <.001 .99 <.001 .99 <.001 .99 <.001
Step time (unaffected) .99 <.001 .99 <.001 .99 <.001 .99 <.001
Stride length .99 <.001 .99 <.001 .99 <.001 .99 <.001
Velocity .99 <.001 .94 <.001 .99 <.001 .99 <.001
Cadence .99 <.001 .97 <.001 .99 <.001 .99 <.001

Note. Statistical significance was tested by the Pearson correlation coefficient (rp) at p < .05.

TABLE 3 Means and standard deviations of gait parameters from two motion analysis systems (n = 20)

Preferred speed Fast speed

Gait parameter Mean ± SD FDM mean ± SD p value Mean ± SD FDM mean ± SD p value
Step length (affected; cm) 37.53 ± 9.73 37.62 ± 1.00 .53 43.53 ± 11.57 43.44 ± 11.74 .37
Step length (unaffected; cm) 36.68 ± 12.43 36.76 ± 12.47 .43 41.63 ± 13.57 41.82 ± 13.48 .15
Step time (affected; s) 0.75 ± 0.24 0.75 ± 0.24 .49 0.66 ± 0.17 0.66 ± 0.17 .61
Step time (unaffected; s) 0.62 ± 0.11 0.63 ± 0.11 .10 0.56 ± 0.11 0.56 ± 0.11 .30
Stride length (cm) 74.21 ± 21.17 74.38 ± 21.48 .36 85.16 ± 23.76 85.25 ± 23.80 .65
Velocity (m/s) 0.58 ± 0.23 0.58 ± 0.23 .52 0.75 ± 0.29 0.75 ± 0.30 .27
Cadence (steps/min) 87.11 ± 26.48 86.90 ± 26.52 .20 103.12 ± 19.66 102.77 ± 19.56 .06

Note. Statistical significance was tested by the paired t test at p < .05.
Abbreviation: FDM, Force Distribution Measurement.

TABLE 4 Intratester reliability of a novice for the video-based system (n = 20)

Preferred speed Fast speed

Gait parameter ICC3,1 95% CI SEM p value ICC3,1 95% CI SEM p value
Step length (affected; cm) 0.98 [0.95, 0.99] 1.98 <.001 0.98 [0.94, 0.99] 2.49 <.001
Step length (unaffected; cm) 0.98 [0.96, 0.99] 2.27 <.001 0.99 [0.98, 0.99] 1.74 <.001
Step time (affected; s) 0.95 [0.88, 0.98] 0.06 <.001 0.98 [0.95, 0.99] 0.03 <.001
Step time (unaffected; s) 0.91 [0.76, 0.96] 0.04 <.001 0.97 [0.92, 0.99] 0.03 <.001
Stride length (cm) 0.99 [0.96, 0.99] 3.74 <.001 0.99 [0.97, 0.99] 3.36 <.001
Velocity (m/s) 0.98 [0.96, 0.98] 0.04 <.001 0.99 [0.96, 0.99] 0.05 <.001
Cadence (steps/min) 0.94 [0.85, 0.98] 5.38 <.001 0.97 [0.92, 0.99] 4.62 <.001

Note. Statistical significance was tested by ICC at p < .05.

Abbreviations: CI, confidence interval; ICC, intraclass correlation coefficient; SEM, standard error of measurement.

compared with 61.5 cm for the video-based system), step period (0.73 in the Cutlip et al. (2000) study. The healthy individuals in the Cutlip
s compared with 0.72 s), stride velocity (91.3 cm/s compared with et al. (2000) study walked with higher speed of 1.79 m/s compared
84.3 cm/s), stance duration (0.93 s compared with 0.95 s), and swing with the stroke participants in our study who walked at 0.75 m/s. So
duration (0.52 s compared with 0.51 s) at a slow speed of 0.85 m/s the frequency of the selected tool should cover frequency of the test-
but showed significant differences of step length (81.2 cm compared ing movement to capture the motion precisely. The discrepancy of
with 76.7 cm) and stride velocity (191.0 cm/s compared with 176.3 spatial parameters with increased speed might be due to the sensitiv-
cm/s) when speed of walking is increased to 1.79 m/s. The differences ity of the pressure transducer and the method of calibration that was
demonstrated at the fast-speed walking in healthy individuals might not mentioned in the Cutlip et al. (2000) study. In contrast, our study
be due to the insufficient sampling rate of 60 Hz of the video camera had two calibrated rulers adhered along the middle part of the 3-m
6 of 7 AUNG ET AL.

FDM walkway, which improved the accuracy of the video data. They the increasing distance of camera to the subjects and zooming in to
were attached parallel to the left and right footpaths to ensure the the required capture size. Moreover, it can be minimized by setting
accuracy of walking data of two legs. the camera perpendicular to the calibrated plane. Another error may
A previous study investigated concurrent validity and reliability of relate to the process of identifying gait events.
the GAITRite system against the validated paper-and-pencil and Individuals with stroke who had trouble with foot drag or severe
video-based methods on the spatio-temporal gait parameters in one foot drop and used the orthotic devices such as foot flap or ankle–
healthy young woman walking at different speeds and degrees of step foot orthotic were excluded from the study, caused by the gait identi-
symmetries (McDonough et al., 2001). Excellent correlations of the fication problem of FDM. Therefore, the utilization of the video-based
right (ICC = 0.97) and left (ICC = 0.99) step lengths between the system in patients with stroke may be limited by this issue.
paper-and-pencil and GAITRite systems whereas fair to good correla- The present study investigated the concurrent validity and
tions (ICC = 0.44 for the right step length and ICC = 0.85 for the left intratester reliability only on the spatio-temporal parameters. For
step length) between the video-based and GAITRite systems were more clinical usage, further study should investigate the other param-
found. The reason for the lower correlation for the right step length eters such as lower extremity joint angle and velocity. Moreover, fur-
may be caused by the error from the video camera located at the left ther study should determine the validity of the video-based system
side of the participant. Thus, higher correlation was found in the left with the other standard motion assessment systems such as 3D
step length. This source of error may relate with the video camera's motion analysis system on kinematics data in individuals with stroke
depth of field. In general laboratory configuration, the video camera may be needed. Another limitation of this study was that we did not
was usually located to one side of the participant, which leads to the investigate the intertester reliability. It is important to obtain a high
inaccuracy of data on another side. For our study, the two calibrated degree of agreement among different raters, especially in a multisite
rulers placed at the middle part of the electronic walkway and parallel study. So further study should focus on the intertester reliability. In
to the left and right footpaths on a contrast background improved the addition, gait in other types of neurological conditions should be
identification of the gait events during the gait cycle. This might be tested for more generalizability.
the reason why no difference in data was seen between the left and
right sides in our study when compared with the McDonough et al. 4.2 | Clinical implications
(2001) study.
A video-based system is simple to set up, portable, and low cost. The
For our study, the excellent intratester reliability was found in a
time spent for setup and extraction processes are only 10 min per par-
novice tester when using a video-based system to capture the spatio-
ticipant and the time of novice training for the overall processes is 2
temporal gait parameters in individuals with stroke. This level of reli-
weeks. According to the valid and reliable data demonstrated in this
ability corresponded to the previous studies that reported the
study, the use of this system is practically in clinical approaches to
intratester reliability of.99 for spatio-temporal, .94–.96 for kinematic,
evaluate and reevaluate gait disturbances in individuals with stroke.
and .73–.95 for tibial inclination in healthy subjects (Soda, Carta, For-
mica, & Guglielmelli, 2009; Ugbolue et al., 2013). For data calculation,
AC KNOWLEDG EME NT S
the step length and step time were manually extracted from the
recorded video, and the other parameters were calculated. The time This work was funded by the Norway Scholarship (Mahidol-Norway
used in this process for our novice tester was around 10 min for each Capacity Building Initiative for ASEAN) scholarship. We would like to
participant, depending on how hard to identify the gait events. With thank all individuals with stroke, physical therapists who assisted in
20 participants, the overall time spent in tracking and reporting pro- this study, and Ms Pavika Poramapornpilas for proofreading the
cesses for the video-based system was around 200 min. manuscript.
In conclusion, the video-based system provides applicability in
clinical service, instead of using advanced motion-capture methods. ET HICAL APPROVAL
However, this approach may be challenged by some instrumental fac-
The study was approved by the institutional research ethical commit-
tors such as high-quality video resolution, requirements of calibration
tee (COA: MU-CIRB 2017/178.1010).
tool, and process of calculation (Damsted, Nielsen, & Larsen, 2015).

CONFLIC T OF INT ER E ST
4.1 | Limitations and further study
The authors declare no conflict of interest.
By using the video-based system, there are different sources of errors
that can occur. First, the parallax error can result from the subject's
AUTHOR CONTRIBU TIONS
motions moving away from the optical axis of the camera. This can be
minimized by aligning the optical axis of the camera with the central Conception or design of the work was done by Bovonsunthonchai,
part of the motion and zooming to capture the interesting motion. Aung, Hiengkaew, and Tretriluxana. Data collection was carried out
Second, the perspective error can result from the part of movement by Aung, Rojasavastera, and Pheung-Phrarattanatrai. Data analysis
segments moving out of the calibrated plane. It can be prevented by and interpretation were performed by Bovonsunthonchai and Aung.
AUNG ET AL. 7 of 7

Drafting of the article was done by Bovonsunthonchai and Aung. Crit- with a single accelerometer in the laboratory and community: A feasi-
ical revision of the article was made by Bovonsunthonchai and Aung. bility, validity and reliability study. Journal of Neuroengineering and
Rehabilitation, 14(1), 130. https://doi.org/10.1186/s12984-017-
Final approval of the version to be published was given by
0341-z
Bovonsunthonchai. Munro, A., Herrington, L., & Carolan, M. (2012). Reliability of
2-dimensional video assessment of frontal-plane dynamic knee valgus
ORCID during common athletic screening tasks. Journal of Sport Rehabilitation,
21(1), 7–11. https://doi.org/10.1123/jsr.21.1.7
Sunee Bovonsunthonchai https://orcid.org/0000-0003-1044-3174 Norris, B. S., & Olson, S. L. (2011). Concurrent validity and reliability of
two-dimensional video analysis of hip and knee joint motion during
Vimonwan Hiengkaew https://orcid.org/0000-0002-0842-6976
mechanical lifting. Physiotherapy Theory and Practice, 27(7), 521–530.
Jarugool Tretriluxana https://orcid.org/0000-0002-5303-8569 https://doi.org/10.3109/09593985.2010.533745
Portney, L. G., & Watkins, M. P. (2015). Foundations of clinical research:
Applications to practice (3rd ed.). Philadelphia: F.A. Davis Company.
RE FE R ENC E S Sekiya, N., Nagasaki, H., Ito, H., & Furuna, T. (1997). Optimal walking in
terms of variability in step length. The Journal of Orthopaedic and
Baker, R., Esquenazi, A., Benedetti, M. G., & Desloovere, K. (2016). Gait Sports Physical Therapy, 26(5), 266–272. https://doi.org/10.2519/
analysis: Clinical facts. European Journal of Physical and Rehabilitation jospt.1997.26.5.266
Medicine, 52(4), 560–574. Soda, P., Carta, A., Formica, D., & Guglielmelli, E. (2009). A low-cost video-
Balaban, B., & Tok, F. (2014). Gait disturbances in patients with stroke. PM based tool for clinical gait analysis. Paper presented at the Engineering
& R : The Journal of Injury, Function, and Rehabilitation, 6(7), 635–642. in Medicine and Biology Society, 2009. EMBC 2009. Annual Interna-
https://doi.org/10.1016/j.pmrj.2013.12.017 tional Conference of the IEEE.
Bilney, B., Morris, M., & Webster, K. (2003). Concurrent related validity of Stratford, P. W., & Goldsmith, C. H. (1997). Use of the standard error as a
the GAITRite walkway system for quantification of the spatial and reliability index of interest: An applied example using elbow flexor
temporal parameters of gait. Gait & Posture, 17(1), 68–74. https://doi. strength data. Physical Therapy, 77(7), 745–750. https://doi.org/10.
org/10.1016/S0966-6362(02)00053-X 1093/ptj/77.7.745
Bovonsunthonchai, S., Witthiwej, T., Ngamsombat, C., Sathornsumetee, S., Tanikawa, H., Ohtsuka, K., Mukaino, M., Inagaki, K., Matsuda, F., Teranishi,
Vachalathiti, R., Muangpaisan, W., … Yangyoo, P. (2018). Effect of spi- T., … Saitoh, E. (2016). Quantitative assessment of retropulsion of the
nal tap test on the performance of sit-to-stand, walking, and turning in hip, excessive hip external rotation, and excessive lateral shift of the
patients with idiopathic normal pressure hydrocephalus. Nagoya Jour- trunk over the unaffected side in hemiplegia using three-dimensional
nal of Medical Science, 80(1), 53–60. https://doi.org/10.18999/nagjms. treadmill gait analysis. Topics in Stroke Rehabilitation, 23(5), 311–317.
80.1.53 https://doi.org/10.1080/10749357.2016.1156361
Cutlip, R. G., Mancinelli, C., Huber, F., & DiPasquale, J. (2000). Evaluation Toro, B., Nester, C. J., & Farren, P. C. (2003). The status of gait assessment
of an instrumented walkway for measurement of the kinematic param- among physiotherapists in the United Kingdom. Archives of Physical
eters of gait. Gait & Posture, 12(2), 134–138. https://doi.org/10.1016/ Medicine and Rehabilitation, 84(12), 1878–1884. https://doi.org/10.
S0966-6362(00)00062-X 1016/S0003-9993(03)00482-9
Damsted, C., Nielsen, R. O., & Larsen, L. H. (2015). Reliability of video- Ugbolue, U. C., Papi, E., Kaliarntas, K. T., Kerr, A., Earl, L., Pomeroy, V. M.,
based quantification of the knee- and hip angle at foot strike during & Rowe, P. J. (2013). The evaluation of an inexpensive, 2D, video
running. International Journal of Sports Physical Therapy, 10(2), based gait assessment system for clinical use. Gait & Posture, 38(3),
147–154. 483–489. https://doi.org/10.1016/j.gaitpost.2013.01.018
Giacomozzi, C. (2010). Appropriateness of plantar pressure measurement Wall, J. C., & Scarbrough, J. (1997). Use of a multimemory stopwatch to
devices: A comparative technical assessment. Gait & Posture, 32(1), measure the temporal gait parameters. The Journal of Orthopaedic and
141–144. https://doi.org/10.1016/j.gaitpost.2010.03.014 Sports Physical Therapy, 25(4), 277–281. https://doi.org/10.2519/
Henriksen, M., Lund, H., Moe-Nilssen, R., Bliddal, H., & Danneskiod- jospt.1997.25.4.277
Samsoe, B. (2004). Test–retest reliability of trunk accelerometric gait Wollesen, B., Voelcker-Rehage, C., Willer, J., Zech, A., & Mattes, K. (2015).
analysis. Gait & Posture, 19(3), 288–297. https://doi.org/10.1016/ Feasibility study of dual-task-managing training to improve gait perfor-
S0966-6362(03)00069-9 mance of older adults. Aging Clinical and Experimental Research, 27(4),
Kalron, A., Dvir, Z., Givon, U., Baransi, H., & Achiron, A. (2014). Gait and 447–455. https://doi.org/10.1007/s40520-014-0301-4
jogging parameters in people with minimally impaired multiple sclero- World Health Organization (2010). Global recommendations.
sis. Gait & Posture, 39(1), 297–302. https://doi.org/10.1016/j.gaitpost. Retrieved from http://www.who.int/dietphysicalactivity/factsheet_
2013.07.124 recommendations/en/
Krebs, D. E., Edelstein, J. E., & Fishman, S. (1985). Reliability of observa-
tional kinematic gait analysis. Physical Therapy, 65(7), 1027–1033.
https://doi.org/10.1093/ptj/65.7.1027
Landis, J. R., & Koch, G. G. (1977). The measurement of observer agree-
How to cite this article: Aung N, Bovonsunthonchai S,
ment for categorical data. Biometrics, 33(1), 159–174. https://doi.org/
10.2307/2529310 Hiengkaew V, Tretriluxana J, Rojasavastera R, Pheung-
McDonough, A. L., Batavia, M., Chen, F. C., Kwon, S., & Ziai, J. (2001). The Phrarattanatrai A. Concurrent validity and intratester reliability
validity and reliability of the GAITRite system's measurements: A pre- of the video-based system for measuring gait poststroke.
liminary evaluation. Archives of Physical Medicine and Rehabilitation,
Physiother Res Int. 2019;e1803. https://doi.org/10.1002/pri.
82(3), 419–425. https://doi.org/10.1053/apmr.2001.19778
Moore, S. A., Hickey, A., Lord, S., Del Din, S., Godfrey, A., & Rochester, L. 1803
(2017). Comprehensive measurement of stroke gait characteristics