While prior observer training standardizes criteria for immediate cross-sectional evaluations, the protocol's high sensitivity to measurement variability and unconstrained motor execution limits its clinical application for longitudinal tracking across different sessions.
Key Findings
Results
Inter-rater reliability for sagittal plane joint angles ranged from moderate to good for most variables, except for ankle angle during loading response.
Inter-rater ICC values ranged from 0.628 to 0.789 across most joint variables
Two independent raters analyzed the same video recordings of sagittal hip, knee, and ankle angles
Analysis was performed using Kinovea® software on 60 fps smartphone video
Results
Between-session intra-rater reliability was generally lower than inter-rater reliability, ranging from poor to moderate.
Between-session intra-rater ICC values ranged from 0.278 to 0.580
Lowest intra-rater ICC values were observed at the hip (ICC = 0.278) and ankle (ICC = 0.391) during loading response
One rater repeated measurements seven days after the initial assessment
The range of intra-rater reliability was classified as poor to moderate
Results
Minimal Detectable Change (MDC95) values were larger for intra-rater assessments than for inter-rater assessments.
MDC95 values ranged from 7.07° to 8.95° for inter-rater assessments
MDC95 values ranged from 7.40° to 16.55° for intra-rater assessments
Bland-Altman analysis displayed wide limits of agreement longitudinally
The wider MDC95 range for intra-rater assessments indicates greater measurement variability across sessions
Methods
The study was conducted in an unconstrained, non-laboratory corridor environment with healthy young adults walking at self-selected speed.
Forty-five healthy young adults participated in the study
Walking took place in a non-laboratory corridor setting
Participants walked at a self-selected speed
Video was recorded at 60 frames per second using a smartphone
The protocol was specifically designed to assess lower limb sagittal kinematics outside of traditional laboratory constraints
Conclusions
Between-session intra-rater variability was attributed to evaluator subjectivity in manual frame selection and landmark digitization over time, rather than biological gait fluctuations.
The absence of physical markers was identified as a key factor contributing to measurement variability across sessions
Manual frame selection and landmark digitization were identified as sources of evaluator subjectivity
The authors distinguished between measurement error and true biological gait variation
Traditional 3D gait analysis with physical markers was implicitly contrasted as a more stable reference approach
Conclusions
The markerless 2D video protocol was considered suitable for immediate cross-sectional evaluations but not for longitudinal tracking across sessions.
Prior observer training was noted to standardize criteria for cross-sectional use
The protocol's high sensitivity to measurement variability limits longitudinal clinical application
Unconstrained motor execution was cited as an additional factor limiting longitudinal use
The authors concluded the method is unsuitable for tracking changes across different sessions in clinical settings
What This Means
This research suggests that using a smartphone camera and free software (Kinovea®) to measure how people move their hips, knees, and ankles while walking — without any physical markers attached to the body — produces reasonably consistent results when two different trained observers analyze the same video at the same time. The agreement between two raters was moderate to good for most joints, though ankle measurements during the moment the foot first contacts the ground were less reliable. The study involved 45 healthy young adults walking in a hallway at their own pace, simulating a real-world rather than laboratory setting.
However, when the same observer tried to repeat their measurements one week later on the same videos, consistency dropped considerably, falling into the poor-to-moderate range for most joints. The researchers attributed this not to actual changes in how people walk, but to the difficulty of consistently choosing the exact same video frame and identifying the same body landmarks by eye when no physical markers are present. The margin of error for detecting real change (called Minimal Detectable Change) was also substantially larger across sessions than between two raters looking at the same footage simultaneously — meaning small changes in a patient's gait over time could easily be lost in measurement noise.
This research suggests that this low-cost, markerless smartphone approach may be useful for taking a one-time snapshot of a person's walking pattern — for example, comparing two people or assessing a group at a single point in time — as long as the analysts are trained and work together. However, it is likely not reliable enough to track whether an individual patient's gait is improving or worsening over weeks or months, which is a common need in rehabilitation and clinical care. Future work would need to address the subjectivity of manual landmark identification, possibly through automated pose-estimation software, to improve longitudinal reliability.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Pérez-de la Cruz S, Molina-Rueda F, Fernández-González P. (2026). Reliability of Markerless 2D Video Analysis for Calculating Gait Joint Angles in Uncontrolled Environments.. Sensors (Basel, Switzerland). https://doi.org/10.3390/s26196229