Comparison of markerless and marker-based motion analysis accounting for differences in local reference frame orientation

Markerless motion analysis offers a promising alternative to traditional marker-based motion capture. Differences between the kinematic signals obtained with these systems have been reported, but the hypothesis that these differences stem from reference frame alignment inconsistencies has not been systematically assessed. This study aims to apply a reference frame alignment method (REFRAME) to markerless captured knee kinematic signals to determine if inconsistencies with marker-based results are related to differences in local frame orientation. Ten healthy subjects (5/5 M/F, age 26.8 f 5.6 years, 21.5 f 1.8 kg/m2) were recruited to perform five walking trials each. Data was simultaneously recorded using eight video cameras (Miqus, Qualisys) for the markerless capture system (Theia3D), and twenty-four cameras (Oqus and Arqus, Qualisys) for the marker-based system (Qualisys). Both sets of processed data were exported for kinematic analysis in Visual3D (C-Motion). Implementing the REFRAME approach optimized the markerless kinematic signals towards the marker-based system, resulting in an average rotation of the tibia coordinate system of 13.8 degrees f 3.1 degrees and 4.9 degrees f 3.6 degrees around the y-and z-axes, respectively, and 5.3 degrees f 2.8 degrees, 5.8 degrees f 2.4 degrees, and 7.9 degrees f 4.9 degrees for the femur coordinate system around the x-, y-, and z-axes, respectively. Post-REFRAME results showed improved agreement between markerless and marker-based data in all three planes. The root-mean-square error decreased from 3.9 degrees f 1.5 degrees to 1.7 degrees f 0.2 degrees (sagittal plane), from 6.1 degrees f 1.3 degrees to 1.7 degrees f 0.3 degrees (frontal plane), and from 10.2 degrees f 2.8 degrees to 2.5 degrees f 0.5 degrees (transverse plane). These findings indicate that most of the differences between markerless and marker-based data are likely due to inconsistencies in local frame orientations, suggesting that markerless kinematic signals represent fundamentally similar underlying motion waveforms.