-This observational study aims to compare gait analysis performed using pose estimation algorithms with inertial measurement unit (IMU)-based gait analysis in older adults. Additionally, it aims to determine the reliability of gait analysis using pose estimation algorithms in this population.

The main questions it aims to answer are:

1. Are gait analyses using pose estimation algorithms consistent with those performed using an inertial measurement unit-based system in older adults?
2. Are gait analyses using pose estimation algorithms reliable in older adults?

Participants will take part in two measurement sessions. In the first session, they will be evaluated for inclusion criteria and general health status, and will complete gait analysis using both G-Walk (BTS Bioengineering) IMU sensors and a standard video camera simultaneously. In the second session, scheduled 1-3 days later, participants will perform only the 4-meter walking test, which will be recorded by video for pose estimation analysis.

The primary objective of this observational study is to compare spatio-temporal gait parameters obtained from 2D pose estimation algorithms with those measured by inertial measurement units (IMUs) in older adults. A secondary aim is to evaluate the test-retest reliability of pose estimation-based gait analysis within this population.

Study Type and Sample Size This is an observational study. Based on clinical guidelines and recent instrumental research comparing validity and measurement methods in the literature, the sample size is determined to be at least 30 participants. Volunteers meeting the inclusion criteria will be recruited from patient relatives visiting the Faculty of Physical Therapy and Rehabilitation at Dokuz Eylul University for treatment.

Data Source and Collection No external data sources will be used. Assessments will be conducted via on-site visits. Any incomplete assessment will be considered as missing data. All collected data will be anonymized and stored on a password-protected institutional server. Access will be restricted to authorized research personnel.

Procedure After collecting sociodemographic data, medical history, and past medical records, participants will be screened using the Mini-Mental State Examination (MMSE) and the Timed Up and Go (TUG) test to determine eligibility. For the TUG test, participants will be asked to stand up from an armchair with armrests, walk a distance of 3 meters at a normal pace, turn around, return, and sit down. The time taken will be recorded in seconds. The TUG and MMSE test will be used to screen for basic mobility and functional balance, and to ensure that participants have sufficient physical and mental ability to complete gait trials safely.

In the primary assessment, all participants will undergo a simultaneous 4-meter walk test recorded using both G-Walk (BTS Bioengineering) IMU sensors and a standard video camera. The videos collected simultaneously with the G-Walk measurements will be processed with pose estimation algorithms for gait analysis to extract spatio-temporal gait parameters. To assess test-retest reliability of the pose estimation-based gait analysis (YOLO V.11-Pose Estimation), the same 4-meter walking test using only video recording will be repeated 1 to 3 days later in a subset of participants. Missing data will be documented and addressed accordingly.

IMU-Based Gait Analysis (G-WALK) Inertial measurement unit (IMU)-based gait analysis will be conducted using the BTS G-WALK system, a validated tool for assessing dynamic spatiotemporal gait parameters. Following the recording of demographic data (age, sex, height, weight, and shoe size), the device will be affixed at the level of the S1 vertebra using an adjustable belt. Participants will be instructed to walk a 4-meter straight path marked on the floor. The system will compute the following gait parameters: gait speed (m/s), cadence (steps/min), step length (m), step time (s), stance phase (% of gait cycle), swing phase (% of gait cycle), double support phase (% of gait cycle), single support phase (% of gait cycle).

Gait Analysis via Pose Estimation Algorithms Synchronized video recordings taken concurrently with G-WALK assessments will be used for pose estimation-based gait analysis. Videos will be trimmed to align with the temporal segment corresponding to the G-WALK data. The pose estimation algorithm used will be YOLOv11X-Pose, with a confidence threshold set at 0.25. This custom-built algorithm detects and tracks 17 key joint landmarks on the human body in 2D during walking sequences. The system will extract the same spatiotemporal gait parameters as the IMU-based method (speed, cadence, step length, etc.) by applying a custom algorithm to analyze dynamic joint trajectories. This standardization enables direct comparison between the two measurement methods. The extracted gait features will include: gait speed (m/s), cadence (steps/min), step length (m), step time (s), stance phase (% of gait cycle), swing phase (% of gait cycle), double support phase (% of gait cycle), single support phase (% of gait cycle).

Statistical Analysis Software All analyses will be conducted using SPSS version 27.0 for Windows.

Descriptive Statistics Data will be presented as frequencies. If parametric assumptions are met, means and standard deviations will be reported; otherwise, medians and interquartile ranges will be used.

Comparative Statistics Gait metrics obtained via inertial measurement unit (IMU)-based gait analysis and pose estimation algorithms will be compared. If parametric assumptions are met, paired t-tests will be used; otherwise, Wilcoxon signed-rank tests will be performed. Agreement between measurement methods will be evaluated using Bland-Altman plots and regression analysis.

Reliability Statistics Test-retest reliability of the pose estimation algorithm will be assessed. A subgroup of 30 randomly selected participants will undergo a second 4-meter walk video recording 1 to 3 days after the initial assessment. Agreement between test and retest measurements will be analyzed using Intraclass Correlation Coefficients (ICC).

Handling of Missing Data If participants withdraw at any point or fail to complete assessment procedures properly, data from those participants will be excluded from further analysis.

Inclusion Criteria:

• Ability to walk independently
• Mini-Mental State Examination (MMSE) score ≤ 23/30

Exclusion Criteria:

• Inability to walk independently for more than 20 meters without assistive devices
• Inability to understand or comply with test instructions
• Lack of consent to participate
• Presence of severe neurological, musculoskeletal, cardiac, or psychological disorders

Community dwelling older adults aged above 65

Since gait assessment of older adults using video analysis has been completed, data protection regulations in our country, particularly regarding video data, are strictly followed in accordance with the Personal Data Protection Law.