Abstract

Background. Gait analysis has a variety of applications in both research and clinical settings, with marker-based motion capture (MBMC) considered the gold standard for three-dimensional gait assessments. Markerless motion capture (MMC) is a novel alternative to MBMC for gait analysis, using artificial intelligence and deep learning algorithms to analyze human motion. Further research is needed to understand the reliability and repeatability of MMC under varying gait conditions. The primary purpose of this study was to determine the test-retest reliability and repeatability of MMC technology during comfortable and fast walking trials. Our secondary purpose was to determine if MMC can detect the differences in gait kinematics between comfortable and fast walking trials. Methods. Thirty-four healthy adults completed two visits performing a minimum of three walking bouts at both comfortable and fast walking speeds through a MMC data collection area. An intraclass correlation coefficient (ICC(3,1)) and standard error of measurement (SEM) were utilized to assess gait range of motion (ROM) reliability between walking speeds and visits. A 2x2 analysis of variance (ANOVA) was used to compare the hip, knee and ankle angles between speeds and visits (p<0.05). Statistical analysis for repeatability was performed by comparing hip, knee, and ankle angles across the gait cycle via a two-tailed paired sample t-test (p<0.05) using statistical parametric mapping (SPM). SPM was used to assess the percentage of significant differences in the time series waveforms of the hip, knee, and ankle joints. Results. Regarding reliability, all three joints in all three planes at both speeds demonstrated at least a moderate ICC (>0.50), except for the knee frontal plane (comfortable and fast speeds) and the ankle frontal plane (fast speed). Only ankle frontal plane ROM had a main effect of visit. ROM for the hip sagittal, hip frontal, knee sagittal, ankle frontal and ankle transverse planes all had a main effect of speed. Regarding repeatability, SPM analysis showed no significant differences between visits for either the comfortable or the fast walking conditions. Between the speeds during visit 1 there were significant differences in each joint plane for at least one percentage of the gait cycle. Discussion. MMC does seem to be reliable during walking for the hip, knee and ankle ROM for both comfortable and fast speeds. MMC demonstrated strong repeatability for joint angle waveforms for the hip, knee and ankle with no significant differences during the gait cycle in any of the three planes of motion, demonstrating that MMC is reliable and repeatable between visits. Differences in gait kinematics demonstrated that MMC is sensitive enough to detect movement alterations between walking speeds. Conclusion. These findings indicate that MMC is both reliable and repeatable between days and can effectively detect variation between walking speeds.