Clinical validity and reliability of the pose estimation-based system to determine spatio-temporal gait parameters in older adults: A pilot study

Background: The spatio-temporal gait characteristics reflect physical function, fall risk, and rehabilitation processes in older adults. Recent advancements in artificial intelligence have led to the development of more practically accessible gait analysis systems. The purpose of this study was to examine clinical validity and test-retest reliability of the YOLOv11 algorithm-based system for the assessment of spatio-temporal gait parameters in community-dwelling older adults. Methods: Thirty-one community-dwelling older adults (mean Age: 66.94 years, mean BMI: 26.7) were recruited for the study. Participants were instrumented with a 4-meter gait test using the G-walk system, and video recording simultaneously. The collected videos were processed using the YOLOv11 algorithm to extract the spatio-temporal gait parameters. The consistency of the two methods was analysed using the Pearson correlation coefficient ( r ), and agreement was assessed using Bland-Altman Plots. The test-retest reliability was assessed using intra-class correlation coefficients (ICCs). Results: Primary findings revealed that the YOLOv11 algorithm-based system exhibited strong to excellent consistency with the G-walk system for gait parameters such as cadence, gait speed, and stride length, gait cycle duration, and gait phases distribution ( r =0.60-0.95; p<0.001). Moderate agreement was observed between the two systems for all spatio-temporal variables, except for the first double support, which showed poor agreement. Additionally, reliability was excellent for cadence (ICC=0.92) and gait speed (ICC=0.91), good for gait cycle duration (ICC=0.89) and stride length (ICC=0.90), and moderate for gait phase distributions (ICC=0.59-0.61). Conclusions: The findings of this study suggest that the YOLOv11 algorithm-based system can be a potential alternative for gait analysis in older adults, providing a more accessible and affordable option for clinicians and researchers. Future research is warranted due to the cautious interpretation of stance, swing, and single/double support phases. Trial registration: The study protocol was registered with ClinicalTrials.gov (identifier number: NCT07119944).