Two-dimensional video gait analysis: A systematic review of reliability,validity,and best practice considerations

Abstract

Background:

Motion capture systems are widely used to quantify human gait. Two-dimensional (2D) video systems are simple to use, easily accessible, and affordable. However, their performance as compared to other systems (i.e. three-dimensional (3D) gait analysis) is not well established.

Objectives:

This work provides a comprehensive review of design specifications and performance characteristics (validity and reliability) of two-dimensional motion capture systems.

Study design:

Systematic review.

Methods:

A systematic literature search was conducted in three databases from 1990 to 2019 and identified 30 research articles that met the inclusion/exclusion criteria.

Results:

Reliability of measurements of two-dimensional video motion capture was found to vary greatly from poor to excellent. Results relating to validity were also highly variable. Comparisons between the studies were challenging due to differences in protocols, instrumentation, parameters assessed, and analyses performed.

Conclusions:

Variability in performance could be attributed to study design, gait parameters being measured, and technical aspects. The latter includes camera specifications (i.e. resolution and frame rate), setup (i.e. camera position), and analysis software. Given the variability in performance, additional validation testing may be needed for specific applications involving clinical or research-based assessments, including specific patient populations, gait parameters, mobility tasks, and data collection protocols.

Clinical relevance

This review article provides guidance on the application of 2D video gait analysis in a clinical or research setting. While not suitable in all instances, 2D gait analysis has promise in specific applications. Recommendations are provided about the patient populations, gait parameters, mobility tasks, and data collection protocols.

Keywords

Two-dimensional motion analysis clinical videos kinematics gait analysis

Introduction

Quantification of human gait is essential in both research and clinical contexts. Gait analysis involves the measurement of spatiotemporal, kinematic, and kinetic variables, such as stride length, walking velocity, and joint angles. Research involving gait analysis provides fundamental insights into the function associated with neuromusculoskeletal and motor control systems to improve the understanding of various locomotor disorders and advance treatments and development of assistive technologies.^1–3 In clinical settings, gait analysis has been used to diagnose gait deviations and evaluate effects of interventions, both in short- and long-term follow-ups.^4,5

A variety of gait analysis systems exist to meet differing needs of measurement validity and reliability, portability, cost, and ease of use. “No tech” approaches including observational gait analysis are most often used in clinical practice stemming from their ease of use and affordability. However, these methods are highly subjective, and assessments are dependent on the experience and interpretation of the clinician. “High tech” systems, such as three-dimensional (3D) motion capture systems, are considered the gold-standard in gait analysis in terms of accuracy and reliability. 3D systems are typically comprised of camera arrays that track markers on the subject in a pre-calibrated space and these can be readily combined with force plates or electromyography (EMG) systems to capture kinetic data and muscle activation patterns during gait. Benedetti et al. mentions that for certain patient groups, 3D gait analysis can affect decision-making and clinical outcomes. For example, for cerebral palsy and adult brain injuries, it can inform clinical evaluations and surgery planning to improve outcomes.⁶ However, 3D gait analysis methods are expensive and complicated to apply in regular clinical practice, and thus have primarily been restricted to research applications.^7,8 Furthermore, the use of 3D motion capture is limited by time, space, costs, and training required for operation.⁵ Hence, simpler and more cost-effective gait analysis methods are being constantly sought.^9–13 These approaches may forego some of the advanced high-fidelity measurement of instrumented gait laboratories, for simpler and clinically appropriate systems.¹⁴

Two-dimensional (2D) video motion capture dates back to the origins of gait analysis, more than a century ago.¹ With recent advances in digital videography and computing, it is a relatively simple process to acquire and process high resolution videos.^15,16 The use of 2D video analysis ranges from sports medicine to rehabilitation.^17–19 In rehabilitation, uses include evaluating effectiveness of interventions and assistive devices for mobility of patients with musculoskeletal and neurological disorders to characterize lower limb kinetics and kinematics.^20–22 However, the overall performance of 2D gait analysis systems is not well established; this may be a reason for lack of broad utilization of these systems in both research and clinical realms. A recent systematic review of literature found 2D video analysis to be reliable, but reliability was dependent on type of athletic task performed, type of reliability calculated (i.e. inter-rater vs intra-rater), and place of measurement.²³ Yet, the review only focused on selected athletic tasks and not gait. Hence, guidance in the peer-reviewed literature is limited about the most appropriate applications, reliability, and validity of 2D video gait analysis for rehabilitation purposes.

In the context of gait analysis, reliability relates to the magnitude of the measurement error in observed measurements to the inherent variability in the parameter being measured.²⁴ Reliability is typically assessed as intra-rater reliability (i.e. single rater or tester analyzing the same video on different days), intra-session reliability (i.e. analyzing different videos within the same session), inter-rater reliability (i.e. different raters analyzing the same video), and inter-session (test–retest) reliability (i.e. analyzing videos in trials that occur days apart and in different sessions).^25,26 Reliability is typically measured as intra-class correlation coefficients (ICCs) or the coefficient of multiple correlation.^27–29

Validity refers to the level of agreement between two measurement methods. In the case of 2D gait analysis techniques, it may be of interest to determine the level of measurement agreement to 3D systems. While the use of 3D systems for gait measurement is not a standard practice in clinical care due to their complexity as described earlier, 3D systems are considered to be gold-standard in terms of providing accurate and reliable measurements. Studies assessing validity of 2D video gait analysis typically take concurrent measurements using 2D and 3D systems.^{7,9,10,19,30–38} Common statistical analyses of validity include assessing the level of agreement between measurements using Bland–Altman plots, correlation coefficients, and analyses of variance.^{7,19,30,33,36,38} A high level of agreement is indicative of valid absolute measurement of gait parameters.

While individual articles have reported aspects of 2D gait analysis, to the authors’ knowledge, no study has consolidated this information to provide a comprehensive assessment of the performance (i.e. reliability and validity) of 2D video gait analysis systems, as well as factors that may affect their performance. In this regard, the main objective of this article was to assess the performance of 2D gait analysis systems in relation to technical system configuration. The overall goal was to inform and improve application of 2D gait analysis.

Methods

A literature search was conducted in three databases—MEDLINE, Web of Science, and SCOPUS. The search time frame was 1990 to January 2019. Early work was excluded to focus on modern approaches and technologies that are more relevant today. The following keywords were included in the search: gait, locomotion, 2D, two-dimension, 2-dimension, video, analysis, measure, metric, and assess. The search strategy used in one of the included databases, Scopus, is shown in Figure 1. Reliability and validity were not included as search keywords, because many articles did not perform quantitative reliability studies or assess validity of 2D systems. All publications in English were considered. In addition to the results obtained from the database searches, bibliographies of the found papers were reviewed and relevant papers were included. Duplicates were removed and the remaining set of abstracts was reviewed based on the inclusion criteria. The inclusion and exclusion criteria were evaluated by one author and verified by another. The abstract inclusion criteria were (1) studies on human subjects and (2) the use of a video camera in the study methodology. Studies that used video cameras for the purpose of computer vision applications and gait recognition were excluded as the focus of this article was clinical gait analysis. If the abstract was not clear with respect to the inclusion criteria, the full text was evaluated using the abstract inclusion criteria. Next, the remaining full texts were reviewed, and the following inclusion criteria were applied: (1) quantitative reliability or validity studies performed or measures of variation reported; studies that qualitatively described the reliability or validity of their systems were excluded; (2) walking or running trials. Studies that used only qualitative methods of gait assessment (e.g. Edinburgh visual gait score) were excluded. Studies were also excluded if 2D kinematic and spatiotemporal data were obtained from 3D analysis systems or were used for kinetic gait analysis. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline,³⁹ the flow diagram for the identification and selection of the included papers is presented in Figure 2. The PRISMA checklist (Supplemental material) was used to ensure appropriate reporting throughout the article. A risk of bias assessment was not included in this review as it was felt that there was limited suitability given the nature of the studies. Risks of bias assessments, which consider aspects such as sequence randomization and blinding, are better suited for treatment studies, and preferably randomized controlled trials (RCTs), unlike the studies found in this review.

Figure 1.

Keyword search strategy used in Scopus database.

Figure 2.

Study identification and selection flow diagram.

Once the final set of studies was established, data extraction was accomplished by one author (AM) and verified by a second author (AE). In case of conflict between the two authors, the third author (JA) functioned as a tie breaker. The following data were extracted from the final set of articles: the research methodology, planes of movement and measurement, parameters, reliability, validity, limitations, and recommendations for clinical use. The methodology data included participant information, the software used, camera specifications and setup, absolute measures of variation, and reliability study protocols and metrics. Measurements were comprised of spatiotemporal data, joint angles, and planes of movement (i.e. sagittal, frontal, and transverse). For studies that compared 2D methods with 3D methods, measures and other comparison data were extracted to assess validity. For studies that assessed the reliability of 2D methods, intra-rater, inter-rater, intra-session, and inter-session data were extracted. The level of reliability was characterized whereby ICC ⩾ 0.90 was considered as excellent, 0.80 ⩽ ICC < 0.90 as good, 0.70 ⩽ ICC < 0.80 as fair, and ICC <0.70 as poor.¹⁴

Results

The database searches resulted in a total of 501 articles (Medline: 115; Scopus: 230; and Web of Science: 156). In addition, eight papers, including a dissertation, were identified by examining the bibliographies of the searched paper.^{25,30,31,36,38,40–42} In total, 30 studies were included in this systematic review.^{7,9,10,14,19,25,30–38,40–54}

Study design and overall recommendations for clinical use

The methodological details for each included study are presented in Table 1. In summary, the number of participants in each study ranged from 1 to 572 (median: 17.5). The majority of studies analyzed gait of healthy individuals (n = 24), while one analyzed patients with peripheral arterial disease,⁴⁵ two with spastic cerebral palsy,^36,38 one with patellofemoral pain syndrome,⁴⁷ one with forefoot varus deformity,⁴³ and one with anterior cruciate ligament (ACL) rupture.⁴⁴ Parameters were mainly assessed in the sagittal plane (n = 22),^{7,9,10,14,19,25,31,33,34,36–38,41,42,44,45,47–51,54} followed by the frontal plane (n = 6),^{30,32,35,40,43,52} and then both the sagittal and frontal planes (n = 2).^46,53 Most studies assessed walking (n = 22)^{7,9,10,14,19,25,31,33–38,40,41,43–45,48,49,52,54} and the others running (n = 8).^{30,32,42,46,47,50,51,53} The most commonly assessed parameters were knee extension/flexion (n = 10),^{7,14,19,25,31,44–46,49,53} ankle (n = 7),^{25,31,36,44–46,49} and hip extension/flexion angles (n = 6),^{25,36,44,45,49,54} followed by stance duration,^34,35,40 hip adduction angles (n = 3),^30,46,53 step length (n = 3),^48,50,51 gait velocity (n = 3),^7,44,49 trunk angles (n = 3),^36,45,48 and pelvic drop angles (n = 3).^30,32,46

Table 1.

Study methodology characteristics.

Study	Purpose/goal of study	Participant type, No. of participants	Walking/running, speed, treadmill	Marker placement	Camera model, resolution, frequency (Hz)	Camera: distance above ground, distance from walkway	Software used	Study design: 3D system; 2D and 3D: concurrent or separate data collection; reliability protocol	Plane	Gait parameter	Reliability type; reliability metric	Measures of variation
Bencke et al.³³	Validity of 2D video analysis, assess intra-rater reliability	Healthy, 26	Walking, self-selected, N/A	D	ND, ND, 50	ND, ND	ND	8 camera VICON, separate; interval: 4–9 days	Sagittal	MLA angles	Intra-rater; ICC	SD, SEM
Brown et al.⁴³	Intra-session reliability in order to confidently use system for larger goal of study	Forefoot varus deformity, 24	Walking, self-selected, treadmill	D	ND, ND, 60	38 cm, 140 cm from treadmill	FootTrak	N/A; N/A; number of trials: 2	Frontal	Maximum pronation, calcaneal eversion, maximum pronation velocity, time to maximum pronation, total pronation	Intra-session; ICC	N/A
Button et al.⁴⁴	Inter-session and inter-rate reliability in order to confidently use system for larger goal of study	ACL rupture, 27	Walking, self-selected, N/A	ND	D, ND, 50	1 m, 6 m from walkway	Siliconcoach	N/A; N/A; interval: ND, number of raters: ND	Sagittal	Ankle, knee, and hip displacement angles and spatiotemporal variables	Intra-rater, inter-rater; ICC	N/A
Castelli et al.¹⁰	Validity of 2D video analysis	Healthy, 10	Walking (slow, comfortable, and fast), N/A	Markerless	D, 1280 × 720 pixels, 50	ND, ND	Developed system	6 camera VICON T20; concurrent; N/A	Sagittal	Hip, knee, and ankle kinematics, pelvic tilt; Cadence, walking speed, stride time, stride length	N/A	MAE, RMSD
Castelli et al.⁹	Validity of 2D video analysis	Healthy, 9	Walking (slow, normal, and fast), N/A	Markerless	D, 1280 × 720 pixels, 50	ND, ND	Developed system	6 camera VICON T20; concurrent; N/A	Sagittal	Pelvis sagittal rotation, shank sagittal rotation, thigh sagittal rotation, and foot sagittal rotation	N/A	RMSD, % RMSD
Cornwall and McPoil³⁵	Validity of 2D video analysis, assess intra-session reliability	Healthy, 11	Walking, self-selected, N/A	D	D, ND, 60	28 cm, 5 m behind subject	Peak performance 2D	3 camera peak performance analysis; concurrent; number of trials: 5	Frontal	Heel strike angle, time of heel off, maximum pronation, time to maximum pronation, stance phase duration	Intra-session; ICC	SEM, CV
Cornwall and McPoil³⁴	Intra-session reliability in order to confidently use system for larger goal of study	Healthy, 16	Walking, ND, N/A	D	D, ND, 60	28 cm, 5 m from midpoint of walkway	Peak performance technologies	N/A; N/A; number of trials: 4	Sagittal	Rearfoot motion, stance phase duration	Intra-session; ICC	CV, SD
Crowther et al.⁴⁵	Intra-session reliability in order to confidently use system for larger goal of study	Healthy, PAD-IC, TPAD-IC, 11	Walking, 3.2 km/h, treadmill	D	D, ND, 50	ND, 3 m from line of motion	Siliconcoach	N/A; N/A; number of trials: 5	Sagittal	Joint kinematics including displacements, velocities, and acceleration of ankle, hip, knee, and trunk	Intra-session; ICC	N/A
Damsted et al.⁴²	Assess intra-rater and inter-rater reliability	Healthy, 25	Running, self-selected, treadmill	D	D, 512 × 384, 300	86 cm, 1.5 m from treadmill	Kinovea	N/A; N/A; N/A	Sagittal	Hip and knee angles at foot strike	Intra-rater, inter-rater; Bland–Altman 95% LOA	95% LOA, Bland–Altman
Dingenen et al.⁴⁶	Assess inter-session reliability	Healthy, 21	Running, self-selected, N/A, treadmill	D	D, ND, 120	Sagittal: 0.80 m, 1.40 m from treadmill; frontal: 1.05 m, 2 m from end of treadmill	Kinovea	N/A; N/A; interval: 1 week	Sagittal and frontal	Lateral trunk motion, contralateral pelvic drop, femoral adduction, hip adduction, foot and tibia inclination at IC, knee flexion, ankle dorsiflexion during MS	Inter-session; ICC	SEM, SDD
Dingenen et al.³²	Validity of 2D video analysis, assess intra-rater and inter-rater reliability	Healthy, 15	Running, self-selected, N/A	D	D, ND, 50	0.95 m, 4.5 m from force plate	Dartfish 7.0	10 camera VICON T10; concurrent; Interval: 7 days, number of raters: 2	Frontal	Contralateral pelvic drop, femoral adduction, and knee valgus angles	Intra-rater, inter-rater; ICC	SEM, SDD
Esculier et al.⁴⁷	Assess intra-rater and inter-rater reliability	PFP, 66	Running, self-selected, treadmill	D	D, 1920 × 1080, 30	ND, 2 m from treadmill	Tempo tap	N/A; N/A; interval: 72 h; NUMBER of raters: 2	Sagittal	Step rate	Intra-rater, inter-rater, ICC	SEM
Grunt et al.³⁶	Validity of 2D video analysis, assess intra-rater and inter-rater reliability	Spastic CP, 17	Walking, self-selected, N/A	D	ND, ND, ND	ND, ND	MoXie Viewer	Optotrak; concurrent; interval: ND, number of raters: 2	Sagittal	Ankle, shank, knee, hip, pelvis, and trunk	Intra-rater, inter-rater; ICC	SEM, SDD, Bland–Altman 95% LOA
Handžić and Reed³⁷	Validity of 2D video analysis	Healthy, 1	Walking, ND, N/A	D	D, 1280 × 720, 32	25 cm, 2.5 m from walkway	Developed system	8 camera VICON T20S; separate; N/A	Sagittal	Ankle and knee position measurements	N/A	N/A
Hanley et al.¹⁹	Validity of 2D video analysis, assess intra-rater reliability	Healthy, 12	Walking, race walking speed, N/A	D	D, 1280 × 1024, 100	1.10 m, 10 m from line of motion	Simi Motion (manual and automatic tracking)	12 camera Qualisys Oqus 7 3D; concurrent; interval: 48 h	Sagittal	Knee angles	Intra-rater, ICC; 95% LOA; CV	RMS difference (sensitivity analysis)
Larsen et al.³⁸	Validity of 2D video analysis	Spastic CP, 10	Walking, ND, N/A	D	ND, ND, 25	ND, ND	Goniometer on screen	6 camera VICON; concurrent; N/A	Sagittal	Hip, knee, and ankle angles at different points in gait cycle	N/A	Bland–Altman 95% LOA, Pearson’s R
Mannon et al.⁴⁰	Intra-session reliability in order to confidently use system for larger goal of study	Healthy, 25	Walking, ND, N/A	D	D, ND, 60	28 cm, 5 m behind subject	Peak Performance 2D	N/A; N/A; number of trials: 3	Frontal	Rearfoot angle	Intra-session; ICC	SEM
Maslivec et al.⁴⁸	Intra-session reliability in order to confidently use system for larger goal of study	Healthy, 31	Walking, (slow, comfortable, fast), N/A	D	ND, ND, 100	ND, ND	Kinovea	N/A; N/A; number of trials: 3 at three different speeds	Sagittal	Head and trunk flexion angles	Intra-session; ICC	TE (on sensitivity)
Maykut et al.³⁰	Validity of 2D video analysis, assess intra-rater reliability	Healthy, 24	Running, self-selected, treadmill	D	ND, ND, ND	ND, ND	Dartfish	10 camera Motion Analysis Corp.; concurrent; interval: 1 week	Frontal	Peak knee abduction angle, contralateral pelvic drop angle, hip adduction angle during midstance	Intra-rater; ICC	N/A
Nielsen³¹	Validity of 2D video analysis, assess intra-rater reliability	Healthy, 18	Walking, self-selected, N/A	D	D, ND, 25	55 cm, 3 m from center of walkway	HMA Technology	8 camera Qualisys ProReflex; concurrent; interval: ND	Sagittal	Ankle and knee angles	Intra-rater; ICC	N/A
Ostrosky et al.⁴⁹	Intra-rater reliability in order to confidently use system for larger goal of study	Healthy, 60	Walking, self-selected, N/A	D	D, ND, 60	81 cm,6.7 m from walkway	ExpertVision Motion Analysis	N/A; N/A; interval: ND	Sagittal	Peak flexion and extension of ankle, knee, and hip and spatiotemporal variables	Intra-rater; ICC	N/A
Padulo et al.⁵¹	Inter-session reliability in order to confidently use system for larger goal of study	Healthy, 16	Running, various speeds, treadmill	ND	D, ND, 210	1.5 m, 6 m from walkway	Dartfish 5.5Pro	N/A; N/A; interval: ND	Sagittal	Contact time, flight time, step length, step frequency	Inter-session; ICC	SEM
Padulo et al.⁵⁰	Inter-session reliability in order to confidently use system for larger goal of study	Healthy, 65	Running, varied speed for each participant, treadmill	D	D, ND, 210	1.5 m, 6 m from walkway	Dartfish 5.5Pro	N/A; N/A; interval: ND	Sagittal	Contact time, flight time, step length, and step frequency	Inter-session; ICC	SEM
Palad et al.⁵²	Inter-rater reliability in order to confidently use system for larger goal of study	Healthy, 572	Walking, self-selected, N/A	D	D, ND, ND	3 m, end of 10 m walkway	Kinovea	N/A; N/A; number of raters: 3	Frontal	HKA angle deviation	Inter-rater; ICC	N/A
Paul et al.¹⁴	Assess intra-session, inter-session, and inter-rater reliability	Healthy, 10	Walking, 1.0 m/s, treadmill	D	D, ND, 30	ND, 10–12 ft from subject	Dartfish	N/A; N/A; number of raters: 2, number of trials: 5, interval: between 1 and 4 weeks	Sagittal	Knee angles, stride length	Inter-rater, intra-session, inter-session; ICC	CMC, MDC
Reinking et al.⁵³	Assess intra-rater and inter-rater reliability	Healthy, 10	Running, self-selected, treadmill	D	D, ND, 240	ND, ND	Kinovea	N/A; N/A; interval: 1 week, number of raters: 6	Sagittal and frontal	Knee flexion, vertical position, hip adduction, and tibia/shank angles	Intra-rater, inter-rater; ICC	SEM
Ross et al.²⁵	Assess intra-rater, intra-session, and inter-rater reliability	Healthy, 32	Walking, self-selected, N/A	D and markerless	D, 1600 × 1200, 30	58 cm, 127 cm from edge of walkway	GAITRite	N/A; N/A; interval: 6 and 12 weeks; number of raters: 3; number of trials: 2	Sagittal	Ankle, knee, and hip angles	Intra-rater, inter-rater, and intra-session; ICC	SD, SEM, MDC
Saner et al.⁵⁴	Inter-session reliability and minimal detectable change	Healthy, 21	Walking, seven speeds, treadmill	D	D, 1080 pixels, 30	ND, ND	Developed system	N/A; N/A; interval: 1 week	Sagittal	Knee and hip angles	Inter-session; ICC, Bland–Altman plots	MDC
Ugbolue et al.⁷	Validity of 2D video analysis, assess intra-rater and inter-rater reliability	Healthy, 12	Walking, ND, N/A	D	D, 480 × 360, 210	ND, 4 m from edge of walkway	ProTrainer System	12 camera VICON, concurrent; interval: 1 h, number of raters: 3	Sagittal	Ankle and knee angles, hip center location	Intra-rater, inter-rater; ICC	SD, mean differences, Cronbach’s coefficient alpha
Wall and Crosbie⁴¹	Assess intra-rater and inter-rater reliability	Healthy, 5	Walking, (slow, moderate, fast, and self-selected), N/A	ND	D, ND, ND	ND, 5 m from middle of walkway	Hypercard	N/A; N/A; interval: not stated; number of raters: 5 (4 in one analysis, 1 in two other analyses)	Sagittal	Spatiotemporal variables	Intra-rater, inter-rater; ICC	CV

ACL: anterior cruciate ligament; CMC: coefficient of multiple correlation; CP: cerebral palsy; CV: coefficient of variation; dist: distance; HKA: hip–knee–angle; ICC: intra-class correlation coefficient; LOA: limits of agreement; MAE: mean absolute error; MDC: minimum detectible change; MLA: medial longitudinal arch; NA: not applicable; ND: not described; PAD-IC: patients with peripheral arterial disease-intermittent claudication; PFP: patellofemoral pain; RMS: root mean square; RMSD: root mean square deviation; D: stated; SD: standard deviation; SDD: smallest detectable differences; SEM: standard error of mean; TE: typical error; TPAD-IC: patients with treatment for peripheral arterial disease-intermittent claudication.

From the authors’ conclusions and recommendations, 14 studies supported the use of 2D analysis in clinics as an alternative to 3D gait analysis.^{9,10,14,32,33,37,41,42,44,46,47,52–54} Two studies recommended that 2D gait analysis should be used with caution due to interference from movements in other planes³⁶ and low validity.³⁸ In addition, one study stated that 2D video systems are only partially viable for use in a clinical setting, because only a limited set of clinically relevant gait variables can be attained reliably and with validity.³⁰ Two studies reported that 2D analysis does not substitute for motion analysis systems having high resolution and sampling rates in a clinical setting.^33,37 Multiple papers recognized that testing with healthy participants limited application of study findings to patients.^30,31,35,46

Technical and configuration information

Camera and walkway setup

Most of the studies provided details of their setup, including the camera distance from walkway (n = 21)^{7,14,19,25,31,32,34,35,37,40–47,49–52} and camera height (n = 16).^{19,25,31,32,34,35,37,40,42–44,46,49–52} Camera height with respect to the ground ranged from 25 cm to 3 m—a median of 80.5 cm. The camera position relative to the walkway ranged from 1.27 to 10 m in the sagittal plane (median of 3.2 m) and relative to the walkway or participant in the frontal place from 1.4 to 10 m (median of 4.75 m). Note that the distance from the walkway was not a standardized metric throughout the studies with some studies defining this distance from the center of the walkway^31,34,41 and others the edge.^7,25 Eleven studies used treadmills,^{14,30,42,43,45–47,50,51,53,54} while all others used overground walking/running. Treadmills can significantly decrease the effect of “parallax” error in the sagittal plane and perspective error in the frontal plane.⁵⁵

Camera parameters

The camera frequency ranged from 25 to 300 Hz, with the majority from 50 to 60 Hz. Information about camera models, including their resolutions, aperture settings, and light sensitivity, was often not provided; only nine studies stated the camera resolution.^{7,9,10,19,25,37,42,47,54} One study indicated that a low frame rate of 25 Hz was a limitation, because high speed ankle movements were blurred.³⁸ Low frame rate (30 Hz) and camera resolution (1600 × 1200 pixels) affected consistency in choosing phases of the gait cycle for analysis.²⁵ In that study, the researcher attempted to select initial contact in both trials, but due to the camera limitations, initial contact was selected for one trial, while the loading phase was selected in another trial.²⁵ In addition, the low resolution accounted for a limited number of pixels around the identified anatomical landmarks, affecting the reliability of the joint kinematics.²⁵ Pixilation was reported at high frame rates (e.g. 300 Hz) causing blurriness.⁴² Use of sufficient lighting at high velocity activity and shutter speeds was recommended.⁴²

Markers

Markers were placed on anatomical landmarks in most of the studies (n = 24),^{7,14,19,30–38,40,42,43,45–50,52–54} while two used markerless systems (n = 2)^9,10 and one did assessments both with and without markers (n = 1).²⁵ To measure ankle, knee, and hip kinematics in the sagittal plane, markers were typically placed on the iliac crest, greater trochanter, lateral femoral epicondyle, lateral malleolus, and fifth metatarsal head.^25,45,49,54 For the frontal plane, markers were placed on the calcaneus, tibia, and navicular tuberosity to measure kinematics of the foot,^33,40 or on the anterior superior iliac spine (ASIS), midpoint of the tibiofemoral joint, and bisection of the medial and lateral malleoli to measure knee abduction, contralateral pelvic drop, and hip adduction.³⁰ Contralateral pelvic drop angle was described as the “angle between the horizontal line starting at the ASIS of the stance leg and a second line connecting the ASIS of the stance and swing leg.”³²

Software

All but one of the studies named the software they used.³³ Four studies developed their own software and algorithms used for 2D analysis.^9,10,37,54 Most commonly used software were Dartfish (n = 5),^{14,30,32,50,51} Kinovea (n = 5),^{42,46,49,52,53} Peak Performance (n = 3),^34,35,40 and Siliconcoach (n = 2).^44,45 Based on the reliability and validity studies in Table 1, Dartfish and Kinovea were used often in the last decade. The studies that compared 2D and 3D motion capture systems (n = 12) employed 3D systems including the VICON, Qualisys, Peak Performance Analysis, and Motion Analysis Corporation (Cortex) systems.

One study described that the GAITRite 2D analysis software processed landmarks as if there was large movement, while in fact there was minimal movement; this resulted in incorrect joint angle measurements.²⁵ While the HMA Technology software could not be fully validated against 3D systems due significant differences with knee angles at initial contact and during swing phase of gait during highly dynamic movements, it was described as promising.³¹ This was related to difficulties in tracking highly dynamic movements. One study indicated that a limitation of their work was the simple, yet time-consuming manual data processing.⁷

Reliability

Twenty-six studies reported reliability of kinematic and spatiotemporal variables (Table 2),^{19,30–36,40,43–51} with all but one reporting the ICC.⁴² Intra-rater (n = 13),^{7,19,25,30–33,36,41,44,47,49,53} inter-rater (n = 10),^{7,14,25,32,36,41,44,47,52,53} intra-session (n = 8),^{14,25,34,35,40,43,45,48} and inter-session (n = 5)^{14,46,50,51,54} reliability are presented as the ICC. For intra-rater and inter-session reliability, the interval times ranged from 1 h to 12 weeks, and most common was a 1-week interval. For the three inter-session studies that reported the intervals, they ranged from 1 to 4 weeks.^14,46,54 For inter-rater reliability, the number of raters ranged from 2 to 6 (median = 3). All the studies that assessed intra-session reliability stated the number of trials performed from 2 to 5 (median = 3.5) per participant.

Table 2.

Reliability of evaluated studies (ICC).

Gait parameter	Intra-rater		Inter-rater		Intra-session		Inter-session
Gait parameter	Reliability summary	Study and data	Reliability summary	Study and data	Reliability summary	Study and data	Reliability summary	Study and data
Hip extension/flexion angles (sagittal)	Poor to excellent	0.87^{a 44} 0.996 (Ext); 0.990 (Flex)⁴⁹ 0.43–0.92 (w. markers)²⁵ 0.06–0.91 (w/o markers)²⁵ 0.75–0.95^{a 36}	Poor to excellent	~0.60–0.70 (w. markers)²⁵ ~0.25–0.34 (w/o markers)²⁵ 0.67–0.93^{a 36}	Poor to excellent	0.8–0.85^{a 45} ~0.09–-0.82 (w/o markers)²⁵	Fair to excellent	0.75–0.93⁵⁴
Hip adduction angles (frontal)	Poor to excellent	0.951–0.963³⁰ 0.08–0.96⁵³	Good to excellent	0.83–0.99⁵³			Good	0.83–0.87^{a 46}
Knee extension/flexion angles (sagittal)	Poor to excellent	1.00¹⁹ 0.941–0.972⁷ 0.990³¹ 0.75^{a 44} 0.994 (Ext); 0.995 (Flex)⁴⁹ 0.11–0.97 (w. markers)²⁵ 0.15–0.78 (w/o markers)²⁵ 0.54–0.99⁵³ 0.85–0.98^{a 36}	Poor to excellent	0.980–0.992⁷ 0.975¹⁴ ~0.62–0.86 (w. markers)²⁵ ~0.53–0.68 (w/o markers)²⁵ 0.76–1.00⁵³ 0.88–0.97^{a 36}	Poor to excellent	0.911¹⁴ 0.8–0.85^{a 45} ~0.5–0.85 (w/o markers)²⁵	Fair to excellent	0.904¹⁴ 0.75–0.93⁵⁴ 0.87–0.89^{a 46}
Knee abduction angles (frontal)	Excellent	0.955–0.976³⁰
Knee valgus angles (frontal)	Excellent	0.90–0.99³²	Excellent	0.90–0.99³²
Ankle angles (sagittal)	Poor to excellent	0.810³¹ 0.78^{a 44} 0.958 (DF); 0.995 (PF)⁴⁹ 0.17–0.93 (w. markers)²⁵ 0.12–0.93 (w/o markers)²⁵ 0.89–0.98^{a 36}	Poor to excellent	~0.4–0.59 (w. markers)²⁵ ~0.58–0.61 (w/o markers)²⁵ 0.83–0.96^{a 36}	Poor to excellent	0.8–0.85^{a 45} ~0.07–0.8 (w/o markers)²⁵	Fair to excellent	0.75–0.90 (DF)^{a 46}
Pelvic angles (sagittal)	Fair to excellent	0.76–0.92^{a 36}	Poor to good	0.55–0.86^{a 36}
Pelvic drop (frontal)	Excellent	0.958–0.966³⁰ 0.90–0.99³²	Excellent	0.90–0.99³²			Poor to good	0.61–0.80^{a 46}
Tibia inclination angle	Fair to excellent	0.731–0.954⁷	Good to excellent	0.822–0.998⁷			Good to excellent	0.87–0.92^{a 46}
Tibia/shank angle (sagittal)	Poor to excellent	0.62–0.94⁵³ 0.92–0.99^{a 36}	Good to excellent	0.88–0.99⁵³ 0.91–0.99^{a 36}
Gait velocity	Excellent	0.993⁷ 0.969⁴⁹	Excellent	0.997–0.999⁷ 0.99^{a 44}
Stride time	Excellent	0.923⁴⁹ 0.97⁴¹	Poor to good	0.67–0.87⁴¹
Temporal/spatial symmetry	Excellent	0.993⁷
Braking double support	Poor	0.69⁴¹	Poor	0.29–0.63⁴¹
Stride length	Excellent	0.956⁴⁹	Good	0.894¹⁴	Excellent	0.97¹⁴
Single support	Fair	0.78⁴¹	Poor to good	0.41–0.84⁴¹
Thrusting double support	Good	0.82⁴¹	Poor	0.30–0.55⁴¹
Total support	Fair	0.79⁴¹	Poor to fair	0.60–0.79⁴¹
Swing duration	Fair	0.76⁴¹	Poor to fair	0.49–0.79⁴¹
Stance duration					Excellent	0.983³⁵ 0.976⁴⁰ 0.999³⁴
Step rate	Excellent	0.984–0.998⁴⁷	Excellent	0.980⁴⁷
Contact time							Excellent	0.97–0.98⁵⁰ 0.96–0.98⁵¹
Flight time							Excellent	0.96–0.98⁵⁰ 0.96–0.98⁵¹
Step frequency							Excellent	0.96–0.98⁵⁰ 0.95–0.98⁵¹
Step length					Fair to excellent	0.77–0.95⁴⁸	Excellent	0.96–0.98⁵⁰ 0.97–0.98⁵¹
Time to heel off					Excellent	0.959³⁵
Time to maximum pronation					Fair to good	0.839³⁵ 0.78–0.86⁴³
Medial longitudinal arch angles	Excellent	0.95³³
Shoe angle (sagittal)	Good to excellent	0.88–0.99⁵³	Excellent	0.98–1.00⁵³
Heel to COM (sagittal)	Fair to excellent	0.79–0.97⁵³	Excellent	0.93–0.99⁵³
COM height (sagittal)	Good to excellent	0.82–0.98⁵³	Excellent	0.93–0.99⁵³
Femoral adduction (frontal)	Excellent	0.90–0.99³²	Excellent	0.90–0.99³²			Excellent	0.91–0.94^{a 46}
Trunk angles (sagittal)	Poor to good	0.58–0.85^{a 36}	Poor to fair	0.48–0.72^{a 36}	Good to excellent	0.89–0.90 (Flex)⁴⁸ 0.8–0.85^{a 45}
Head flexion (sagittal)					Good to excellent	0.89–0.90⁴⁸
Heel strike angle (sagittal)					Excellent	0.969³⁵
Maximum pronation					Poor to excellent	0.987³⁵ 0.63–0.93^{a 43}
Maximum pronation velocity					Poor to fair	0.58–0.76^{a 43}
Total pronation					Poor to good	0.45–0.83^{a 43}
Calcaneal eversion					Fair to excellent	0.77–0.95^{a 43}
HKA (hip–knee–ankle) angle (frontal)			Good	0.85⁵²
Lateral trunk position							Excellent	0.91–0.92^{a 46}
Foot inclination angles							Excellent	0.98–0.99^{a 46}

DF: dorsiflexion; Ext: extension; Flex: flexion; PF: plantar flexion; COM: center of mass; HKA: hip–knee–ankle.

“~” notation is present because these approximate values were obtained from a bar chart.

The study included participants from patient populations (e.g. spastic cerebral palsy, ACL rupture).

Intra-rater reliability of lower limb joint angles as well as gait velocity and stride time varied across the studies—ranging from poor to excellent. Excellent intra-rater reliability was found for some gait parameters, including pelvic drop angles,^30,32 gait velocity,^7,49 and stride time.^41,49 In the sagittal plane, intra-rater reliability of hip, knee, and ankle angles was poor to excellent.^{25,31,36,44,49,53}

Inter-rater reliability was inconsistent for hip flexion/extension, knee flexion/extension, and ankle angles. Good to excellent inter-rater reliability was reported in studies that measured tibia/shank angles^36,53 and gait velocity.^7,44

The intra-session ICC values for hip, knee, and ankle angles were poor to good.^25,45 For trunk angles, the ICC values were good to excellent.^45,48 For stance duration, intra-session reliability was excellent.^34,35,40

The inter-session reliability for knee flexion/extension angles was fair to excellent.^14,54 Ankle⁴⁶ and hip angle⁵⁴ ICC values could not be compared between studies, because only one study reported these reliability results for each parameter.

Comparing reliability with and without markers²⁵ showed that markers produced higher intra-rater and inter-rater ICC values. Studies that used Kinovea software produced reliability ICC results ranging from poor to excellent.^46,48,52,53 Dartfish software produced ICC reliability results of good to excellent.^{14,30,32,50,51} ICC results for Siliconcoach software was fair to excellent,^44,45 while Peak Performance software showed good to excellent^34,35,40 reliability ICC values. Camera distance from walkway for sagittal plane studies and camera distance from participant/treadmill for frontal plane studies were plotted against reliability, as shown in Figure 3. However, these results should be interpreted with caution as 2D camera calibration can affect the image quality and ultimately the reliability. Camera frame rates are plotted against the study reliability in Figure 4. To compare potential differences in overground and treadmill reliability, ICC values are plotted for both treadmill and overground walking/running studies in Figure 5. Studies with the camera set back at least 6 m from the walkway/participant/treadmill produced good to excellent reliability in the sagittal plane, whereas a trend was not evident for in the frontal plane. In addition, a trend was not evident with frame rate and ICC reliability results. The studies that included patient populations had reliability ICC values that ranged from 0.45 to 0.99.^{36,38,43–45,47} The remaining 24 studies focused on healthy participants had reliability ICC values that ranged from 0.05 to 1. No conclusions can be drawn from this.

Figure 3.

Camera distance from walkway (m) vs reliability (ICC).

Figure 4.

Camera frame rate (Hz) vs reliability (ICC).

Figure 5.

Overground and treadmill vs reliability (ICC).

Validity

Twelve studies compared 2D to 3D motion analysis systems (Table 3).^{7,9,10,19,30–33,35–38} Varied statistical tests and methods were used making it challenging to compare results. Some of the frequently used methods included the Bland–Altman limits of agreement, paired t-tests, and Pearson product correlation coefficients. It is important to note that there is questionable acceptability of the latter two measures to assess agreement.²⁴ When used in isolation, they cannot determine levels of agreement or validity; this should be considered in the interpretation of the results.

Table 3.

Validity including statistical tests, parameters, and results for 2D and 3D comparison.

Study	Comparison methods/statistical tests	Parameter; compare wave forms/discrete parameters	Results
Grunt et al.³⁶	Bland–Altman 95% limits of agreement	Peak ankle dorsiflexion, knee extension, and hip extension during stance and swing phases	Large variation, agreement worst for peak ankle dorsiflexion in swing phase and best for peak knee flexion in swing phase
Larsen et al.³⁸	Bland–Altman with mean differences and 95% limits of agreement, Pearson’s correlation coefficient supplementary	Hip, knee, and ankle angles; at different points in the gait cycle (initial contact, end double support, midstance, start double support, toe off, and midswing)	Knee joint had best agreement
Bencke et al.³³	Paired Student’s t-tests (p < 0.001), Pearson’s correlation coefficient (p < 0.001)	MLA angles; at heel strike with no load and peak ground reaction push force	Significant difference between 2D and 3D MLAD, r = 0.76
Castelli et al.¹⁰	Shape similarity compared (R²), amplitude (A0) and offset (A1)	Hip, knee, and ankle kinematics, pelvic tilt; joint kinematic curves compared	Hip and knee joint kinematics: excellent correlation of 0.96–0.99, ankle kinematics: strong correlation of 0.82–0.97, pelvic tilt: no correlation; A0 amplitude underestimated for hip and ankle joint kinematics, overestimated for knee kinematics, A1 offset was consistent
Castelli et al.⁹	Correlation coefficient (CC), root mean square deviation (RMSD), RMSD normalized to range of motion (%RMSD)	Pelvis sagittal rotation, shank sagittal rotation, thigh sagittal rotation, and foot sagittal rotation; each video frame of gait cycle	Limited errors for foot and shank angular kinematics, greater errors for pelvis. Correlation strong for foot, shank, and thigh. High %RMSD for pelvis angular kinematics.
Cornwall and McPoil³⁵	Pearson product–moment correlation coefficient, paired t-tests, series of multivariate T² tests (α = 0.05)	Motion patterns and stance phase durations compared (Pearson product–moment correlation); variables (paired t-tests); motion patterns at every 2% of the gait cycle from 0% to 60% (T² tests)	r = 0.975 for motion patterns and r = 0.967 for stance duration, t-test: stance duration and heel strike angle are significantly different but time to heel off, maximum pronation, and time to maximum pronation are not significantly different, motion patterns are not statistically different except for first 6% of stance duration
Dingenen et al.³²	One-dimensional parametric mapping (SPM1D) linear regression	2D femoral adduction and hip adduction regressed against 3D hip adduction. 2D contralateral pelvic drop regressed against 3D contralateral pelvic drop and hip adduction. 2D knee valgus regressed against 2D knee abduction	2D femoral adduction and hip adduction significantly negatively related to 3D hip adduction. 2D contralateral pelvic drop significantly positively related to 3D hip adduction and contralateral pelvic drop. No relationship between 2D knee valgus and 2D knee abduction. Results vary over different phases of the gait cycle
Handžić and Reed³⁷	Magnitudes and deviations	Ankle and knee position measurements; wave forms compared	High correspondence of ankle marker X and Z position. Both setups record similar marker position trends, magnitudes, and deviations. Standard deviation increases as marker velocity at foot toe off increases
Hanley et al.¹⁹	ANOVA with post-hoc Tukey tests (α = 0.05), effect size using Cohen’s d	Knee angles—automatic tracking 2D vs 3D and manual tracing 2D vs 3D; at initial contact and midstance	No differences in mean angle at initial contact, but mean angle larger than automatic tracking 2D during mid-stance for 3D system
Maykut et al.³⁰	Pearson product correlation coefficients, Bland–Altman plots	Peak knee abduction angle, contralateral pelvic drop angle, hip adduction angle; during midstance	Peak hip adduction moderately correlated; no significant correlation for contralateral pelvic drop and inconsistent findings for knee abduction. Bland–Altman—no systematic shift between 2D and 3D
Nielsen³¹	Wilcoxon signed-rank tests, Spearman correlation tests, mean difference, and standard deviationBland–Altman for visual relative knee and ankle angles randomized during gait	Comparison in four situations: relative knee and ankle angles randomized during gait, angles at initial contact, angles during stance, and angles during swing	Significant differences between the systems in most cases, aside from knee angles during randomized gait and stance phase; Knee—high correlation; Ankle—moderate to high correlation
Ugbolue et al.⁷	Coefficient of determination (R²), mean difference, linearity expressed as percentage error between 2D and 3D system from 0° to +90° and 0° to −90°, ANOVA (p < 0.05)	Ankle and knee angles, hip center location; at initial contact, foot flat, midstance, terminal contact	Strong correlation R² = 0.999, mean difference very small −0.9° to 0.8°, 0.88 and 1.0, no significant differences between two systems

MLA: medial longitudinal arch; MLAD: medial longitudinal arch deformation; CC: correlation coefficient; RMSD: root mean square deviation.

Two studies used the Bland–Altman limits of agreement to assess validity.^36,38 Both found that the knee angles had the best agreement, especially in the swing phase of the gait cycle.³⁶ Findings overall were variable; in general, moderate to high correlations existed for certain parameters, such as knee and ankle angles,^7,31 but not for others (i.e. ones with small excursions such as pelvic tilt and pelvic drop or in the secondary planes). No significant correlations to 3D systems were found for contralateral pelvic drop, and inconsistent findings were reported for knee abduction in the frontal plane.³⁰ Few articles indicated that the 2D and 3D methods were not statistically different and had strong correlation,^7,35 with the exception of the first 6% of the stance phase in one study.³⁵ Two studies assessed the validity of frontal plane joint kinematics using Dartfish 2D video software when compared to 3D motion analysis.^30,32 Both studies assessed validity of hip adduction angle and contralateral pelvic drop angle using 2D system and correlation coefficient findings were inconsistent. Few studies found no correlation in some measurements, but high correlation in others.^9,10,30 In summary, the results of the review do not support an overall validation of 2D systems; however, it does identify specific gait parameters that can be captured using 2D systems such as knee and ankle angles with validity. Finally, it should be noted that all of the validity studies with the exception of one³⁰ used overground walking rather than treadmill walking, so it was not possible to compare results on this basis.

Discussion

This review aimed to provide information about the performance of 2D gait analysis and guidance about the technical and protocol-related factors. Fatone and Stine¹⁶ recommended that equipment for a good quality video should include a camera, tripod, leveling device, markers to track anatomical landmarks, computer system, and digitizing software. While our review agrees with many of these previous recommendations, it also provides new and important insights based on information extracted from recent published studies. This section highlights these findings and summarizes the overall state of video analyses for gait assessment.

Reliability and validity across studies greatly varied from poor to excellent. Comparisons between the studies were challenging, attributable to variabilities in the protocols, instrumentation, parameters assessed, and analyses performed. Due to many factors, including limited number of gait parameters assessed and studies available, it is difficult to perform meta-analyses to inform which factors influence reliability. As such, our results and recommendations are partly based on descriptive analyses and recommendations provided in the reviewed papers. Variability in results limits a blanket application of 2D gait analysis. While validity and reliability information exists for specific individual parameters, other gait parameters of interest that were not included in this review may need to be further researched and performance must be established. As such, this article provides guidance about important considerations for the technical setup and performance assessments (i.e. what to measure and how to perform reliability and validity assessments).

In a few studies, reliability results were reported to be poor and unacceptable.^{25,36,41,43,46,53} One potential cause was low frame rate and resolution of the videos,²⁵ but this relationship was not clearly evident in our analysis. Nevertheless, frame rates of at least 50–60 Hz, high definition camera resolution, and adequate lighting are recommended to capture high velocity movement without blurring. In terms of software, Kinovea and Dartfish were among the few systems that were used in multiple studies. This may be because Dartfish offers automatic marker tracking, while Kinovea offers semi-automatic tracking. Dartfish produced good to excellent reliability results in the included studies and is recommended for use in future studies. The use of markers is recommended for effective and efficient use of the software, Fatone and Stine¹⁶ also suggest using markers for 2D analysis as manually digitizing without markers is potentially error prone. With marker use, errors may occur during 2D analysis as they do with 3D gait analysis due to soft tissue artifacts and inconsistent marker placement.^6,16 Another potential source of error is related to parallax and perspective, which can be reduced using a treadmill.¹⁶ However, no clear benefits for treadmill gait were found in terms of reliability results and validity could not be compared since only one study used a treadmill.³⁰ In terms of camera height with respect to the ground, parallax is said to be minimized if the camera is perpendicular to the measurement area of interest.¹⁵

For kinematics, knee, hip, and ankle angles in the sagittal plane were most commonly extracted from the videos. This may be for several reasons. Sagittal plane movements of the lower limbs often reveal clinically relevant gait abnormalities.^56,57 In addition, the measurement and extraction of parameters can be achieved more reliably in this plane due to larger and more defined movements.^1,15 Abduction and adduction angles are also deemed important, especially when evaluating the gait of persons with hemiplegia to determine the cause of gait pathologies.⁵⁸ However, measurement of gait parameters from these secondary (i.e. frontal and transverse) planes is more challenging due to smaller magnitudes of the excursions of movements. Another challenge of frontal plane measurement is inconsistency in the image, as subject size increases and decreases in the case of overground walking.¹⁵ Again, the use of a treadmill can eliminate this problem, where feasible. Fatone and Stine¹⁶ have suggested recording in multiple views in order to capture quality clinical videos; however, this may require the development of image processing and video analysis tools capable of consolidating information from multiple images.

Further limitations and recommendations for 2D gait analysis are noted. First, treadmills can eliminate inaccuracies with parallax and frontal plane image size changes, however, few validity studies exist. Hence, more work is needed to assess and compare performance of overground and treadmill protocols. Most studies have focused on self-selected walking velocities with acceptable blurring and image capture quality;¹⁶ however, more research is necessary at higher gait speeds. Also, there are other activities, such as slope and stair negotiation that are part of clinically relevant mobility and physical tasks that need to be considered when evaluating performance of 2D systems.²³ Furthermore, the fact that only a small portion of studies have used patient populations may suggest limited translation or penetration of 2D gait analysis into the clinical realm. This may be, at least in part, related to some of the performance limitations associated with these systems in terms of validity and reliability, as revealed by this review. Nevertheless, validation studies are needed for 2D gait analysis application with specific patient populations of interest. Finally, spatiotemporal measures were assessed to a lesser degree in the reviewed studies compared to kinematics, potentially since they can be more easily acquired using alternative means (i.e. pressure mats). Several spatiotemporal measures including gait speed, stride time, and stride length can be extracted with high reliability, however, validity studies are missing.

One limitation of this review is that the search was limited to three databases, therefore, some related literature may not have been included in this work. However, these are commonly used databases in this area of research. A risk of bias assessment was not formally conducted; however, most of the papers included in this review were not at the level of RCT’s and did not control for factors such as blinding, randomization of sequence, concealment, etc. Due to the diverse nature of the studies, other risk-of-bias criteria such as outcome reporting were also not formally assessed. Finally, research on markerless techniques using depth cameras and machine learning was not the focus of this review. These techniques may offer simple and affordable movement and gait analyses,⁵⁹ however, these techniques are still in their infancy, and have yet to see translation into clinical practice. Furthermore, a limitation of this review is that some validity studies used parameters such as t-test and correlation coefficients. These parameters cannot be used in isolation to determine validity, but these studies were still included in this review.

Conclusion

To be viable in clinical research, certain technical recommendations provided here should be considered relating to camera and analysis software specifications and data collection setups. More research is needed to identify reliable design specifications and protocols for activities other than walking and running and for different speeds. Future work should aim to standardize protocols for 2D gait analysis and further assess reliability and agreement compared to 3D systems for clinically relevant gait parameters and patient populations.

Supplemental Material

10.1177_0309364620921290_Supplemental_material – Supplemental material for Two-dimensional video gait analysis: A systematic review of reliability, validity, and best practice considerations

Supplemental material, 10.1177_0309364620921290_Supplemental_material for Two-dimensional video gait analysis: A systematic review of reliability, validity, and best practice considerations by Alexandria Michelini, Arezoo Eshraghi and Jan Andrysek in Prosthetics and Orthotics International

Footnotes

Acknowledgements

The authors wish to thank Ms Pui-Ying Wong for her invaluable advice on the search strategy and keywords selection as well as Calvin C. Ngan for his support with reviewing and editing the manuscript.

Associate Editor: Andrea Giovanni Cutti

Author contribution

All authors contributed equally in the preparation of this manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Alexandria Michelini

Arezoo Eshraghi

Jan Andrysek

Supplemental material

Supplemental material for this article is available online.

References

Whittle

. Clinical gait analysis: a review. Hum Movement Sci 1996; 15(3): 369–387.

Baker

Esquenazi

Benedetti

, et al. Gait analysis: clinical facts. Eur J Phys Rehabil Med 2016; 52(4): 560–574.

Whittle

. Gait analysis: an introduction. London: Elsevier Science & Technology, 1991.

Harvey

Gorter

. Video gait analysis for ambulatory children with cerebral palsy: why, when, where and how! Gait Posture 2011; 33(3): 501–503.

Simon

. Quantification of human motion: gait analysis—benefits and limitations to its application to clinical problems. J Biomech 2004; 37(12): 1869–1880.

Benedetti

Beghi

De Tanti

, et al. SIAMOC position paper on gait analysis in clinical practice: general requirements, methods and appropriateness. Results of an Italian consensus conference. Gait Posture 2017; 58: 252–260.

Ugbolue

Papi

Kaliarntas

, et al. The evaluation of an inexpensive, 2D, video based gait assessment system for clinical use. Gait Posture 2013; 38(3): 483–489.

Chung

PYM

GYF

. Comparison between an accelerometer and a three-dimensional motion analysis system for the detection of movement. Physiotherapy 2012; 98(3): 256–259.

Castelli

Paolini

Cereatti

, et al. 2D video-based human gait analysis: a novel markerless approach. In: Proceedings of the 20th IMEKO TC4 international symposium on measurements of electrical quantities: research on electrical and electronic measurement for the economic upturn, together with 18th TC4 international workshop on ADC and DCA modeling and testing (IWADC’2014), Benevento, 15–17 September 2014, pp. 760–764. Benevento, Italy: IMEKO-International Measurement Federation Secretariat.

10.

Castelli

Paolini

Cereatti

, et al. A 2D markerless gait analysis methodology: validation on healthy subjects. Comput Math Method M 2015; 2015: 186780.

11.

Goffredo

Carter

Nixon

. 2D markerless gait analysis. In: Vander Sloten

Verdonck

Nyssen

, et al. (eds) 4th European conference of the international federation for medical and biological engineering, vol. 22. Berlin; Heidelberg: Springer, 2009, pp. 67–71.

12.

Kusakunniran

. Extracting gait figures in a video based on markerless motion. In: Proceedings of the 2015 7th international conference on knowledge and systems engineering (KSE) (ed Mérialdo

Le Nguyen

D-D

, et al.), Ho Chi Minh City, Vietnam, 8–10 October 2015, pp. 306–309. New York: IEEE.

13.

Taylor

Miller

Kaufman

. Static and dynamic validation of inertial measurement units. Gait Posture 2017; 57: 80–84.

14.

Paul

Petrizzo

Rizzo

, et al. Feasibility of a cost-effective, video analysis software-based mobility protocol for objective spine kinematics and gait metrics: a proof of concept study. PM R 2015; 7(3): 336–339.

15.

Baker

. Measuring walking: a handbook of clinical gait analysis. London: Mac Keith Press, 2013.

16.

Fatone

Stine

. Capturing quality clinical videos for two-dimensional motion analysis. J Prosthet Orthot 2015; 27(1): 27–32.

17.

Hanley

Bissas

. Analysis of lower limb internal kinetics and electromyography in elite race walking. J Sports Sci 2013; 31(11): 1222–1232.

18.

Dingenen

Malfait

Vanrenterghem

, et al. The reliability and validity of the measurement of lateral trunk motion in two-dimensional video analysis during unipodal functional screening tests in elite female athletes. Phys Ther Sport 2014; 15(2): 117–123.

19.

Hanley

Tucker

Bissas

. Differences between motion capture and video analysis systems in calculating knee angles in elite-standard race walking. J Sports Sci 2018; 36(11): 1250–1255.

20.

Park

Son

Lee

. Exercise-induced muscle fatigue in the unaffected knee joint and its influence on postural control and lower limb kinematics in stroke patients. Neural Regen Res 2017; 12(5): 765–769.

21.

Parvataneni

Olney

Brouwer

. Changes in muscle group work associated with changes in gait speed of persons with stroke. Clin Biomech 2007; 22(7): 813–820.

22.

Olree

Engsberg

Ross

, et al. Changes in synergistic movement patterns after selective dorsal rhizotomy. Dev Med Child Neurol 2000; 42(5): 297–303.

23.

Lopes

TJA

Ferrari

Ioannidis

, et al. Reliability and validity of frontal plane kinematics of the trunk and lower extremity measured with 2-dimensional cameras during athletic tasks: a systematic review with meta-analysis. J Orthop Sports Phys Ther 2018; 48(10): 812–822.

24.

Camomilla

Cereatti

Cutti

, et al. Methodological factors affecting joint moments estimation in clinical gait analysis: a systematic review. Biomed Eng Online 2017; 16(1): 106.

25.

Ross

Rice

Von Behren

, et al. Reliability of sagittal plane hip, knee, and ankle joint angles from a single frame of video data using the GAITRite camera system. Physiother Theory Pract 2015; 31(1): 53–60.

26.

Streiner

Norman

. Reliability. In: Streiner

Norman

Cairney

(eds) Health Measurement Scales: a practical guide to their development and use. Oxford: Oxford University Press, 2014, p. 452.

27.

Koo

. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J Chiropr Med 2016; 15(2): 155–163.

28.

Zaki

Bulgiba

Nordin

, et al. A systematic review of statistical methods used to test for reliability of medical instruments measuring continuous variables. Iran J Basic Med Sci 2013; 16(6): 803–807.

29.

Gulgin

Hoogenboom

. The functional movement screening (fms)^TM: an inter-rater reliability study between raters of varied experience. Int J Sports Phys Ther 2014; 9(1): 14–20.

30.

Maykut

Taylor-Haas

Paterno

, et al. Concurrent validity and reliability of 2d kinematic analysis of frontal plane motion during running. Int J Sports Phys Ther 2015; 10(2): 136–146.

31.

Nielsen

. Comparison of angular measurements by 2D and 3D gait analysis. PhD Dissertation, Jönköping University, Jönköping, 2008.

32.

Dingenen

Staes

Santermans

, et al. Are two-dimensional measured frontal plane angles related to three-dimensional measured kinematic profiles during running? Phys Ther Sport 2018; 29: 84–92.

33.

Bencke

Christiansen

Jensen

, et al. Measuring medial longitudinal arch deformation during gait. A reliability study. Gait Posture 2012; 35(3): 400–404.

34.

Cornwall

McPoil

. The influence of tibialis anterior muscle activity on rearfoot motion during walking. Foot Ankle Int 1994; 15(2): 75–79.

35.

Cornwall

McPoil

. Comparison of 2-dimensional and 3-dimensional rearfoot motion during walking. Clin Biomech 1995; 10(1): 36–40.

36.

Grunt

Van Kampen

van der Krogt

, et al. Reproducibility and validity of video screen measurements of gait in children with spastic cerebral palsy. Gait Posture 2010; 31(4): 489–494.

37.

Handžić

Reed

. Analysis of human stepping dynamics using a wii balance board with a webcam: a comparison study. In: Proceedings of the 8th ACM international conference on PErvasive technologies related to assistive environments (PETRA’15), Corfu, 1–3 July 2015, pp. 1–4. New York: ACM Press.

38.

Larsen

Maanum

Froslie

, et al. Ambulant adults with spastic cerebral palsy: the validity of lower limb joint angle measurements from sagittal video recordings. Gait Posture 2012; 35(2): 186–191.

39.

Moher

Shamseer

Clarke

, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4: 1.

40.

Mannon

Anderson

Cheetham

, et al. A comparison of two motion analysis systems for the measurement of two-dimensional rearfoot motion during walking. Foot Ankle Int 1997; 18(7): 427–431.

41.

Wall

Crosbie

. Temporal gait analysis using slow motion video and a personal computer. Physiotherapy 1997; 83(3): 109–115.

42.

Damsted

Nielsen

Larsen

. Reliability of video-based quantification of the knee- and hip angle at foot strike during running. Int J Sports Phys Ther 2015; 10(2): 147–154.

43.

Brown

Donatelli

Catlin

, et al. The effect of two types of foot orthoses on rearfoot mechanics. J Orthop Sports Phys Ther 1995; 21(5): 258–267.

44.

Button

Van Deursen

Price

. Recovery in functional non-copers following anterior cruciate ligament rupture as detected by gait kinematics. Phys Ther Sport 2008; 9(2): 97–104.

45.

Crowther

Spinks

Leicht

, et al. Effects of a long-term exercise program on lower limb mobility, physiological responses, walking performance, and physical activity levels in patients with peripheral arterial disease. J Vasc Surg 2008; 47(2): 303–309.

46.

Dingenen

Barton

Janssen

, et al. Test-retest reliability of two-dimensional video analysis during running. Phys Ther Sport 2018; 33: 40–47.

47.

Esculier

J-F

Silvini

Bouyer

, et al. Video-based assessment of foot strike pattern and step rate is valid and reliable in runners with patellofemoral pain. Phys Ther Sport 2018; 29: 108–112.

48.

Maslivec

Bampouras

Dewhurst

. Head flexion and different walking speeds do not affect gait stability in older females. Hum Movement Sci 2017; 55: 87–93.

49.

Ostrosky

VanSwearingen

Burdett

, et al. A comparison of gait characteristics in young and old subjects. Phys Ther 1994; 74(7): 637–644; discussion 644–646.

50.

Padulo

Annino

Smith

, et al. Uphill running at iso-efficiency speed. Int J Sports Med 2012; 33(10): 819–823.

51.

Padulo

Annino

Migliaccio

, et al. Kinematics of running at different slopes and speeds. J Strength Cond Res 2012; 26(5): 1331–1339.

52.

Palad

Leaver

McKay

, et al. Knee thrust prevalence and normative hip-knee-ankle angle deviation values among healthy individuals across the lifespan. Osteoarthritis Cartilage 2018; 26(10): 1326–1332.

53.

Reinking

Dugan

Ripple

, et al. Reliability of two-dimensional video-based running gait analysis. Int J Sports Phys Ther 2018; 13(3): 453–461.

54.

Saner

Washabaugh

Krishnan

. Reliable sagittal plane kinematic gait assessments are feasible using low-cost webcam technology. Gait Posture 2017; 56: 19–23.

55.

Kirtley

. Clinical gait analysis: theory and practice. Edinburgh; New York: Elsevier, 2006.

56.

Andrysek

Redekop

Naumann

. Preliminary evaluation of an automatically stance-phase controlled pediatric prosthetic knee joint using quantitative gait analysis. Arch Phys Med Rehabil 2007; 88(4): 464–470.

57.

Rodda

Graham

Carson

, et al. Sagittal gait patterns in spastic diplegia. J Bone Joint Surg Br 2004; 86(2): 251–258.

58.

Churchill

AJG

Halligan

Wade

. RIVCAM: a simple video-based kinematic analysis for clinical disorders of gait. Comput Methods Programs Biomed 2002; 69(3): 197–209.

59.

Poppe

. Vision-based human motion analysis: an overview. Comput Vis Image Underst 2007; 108(1–2): 4–18.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB