(D-125) Markerless and Marker-based Motion Capture Show Comparable Results for Upper Extremity Kinematics in Typically Developing Children during the Box and Blocks Test

Introduction:: Three-dimensional (3D) marker-based motion capture has been considered the “gold-standard” for movement research and assessment. Nonetheless, marker-based motion capture involves extensive preparation, the systems are not portable, data capture must be performed in a controlled lab space, and motion artifact can lead to increased error.¹ Markerless motion capture systems are being introduced as an alternative as they allow for portability, various collection locations, and quick data collection.² Additionally, markerless motion capture could be technology that is introduced to clinicians, specifically in the pediatric space because of the ease of use of the systems. Some validation studies have been completed, however, there is a lack of validation in a pediatric population and upper extremity movements.^3,4 Therefore, this study aimed to compare upper extremity joint kinematics of typically developing children during the Box and Blocks Test (BBT) utilizing marker-based and markerless motion capture.


Materials and Methods:: 21 typically developing children participated in the study through an institutional review board-approved procedure. Children who did not have full upper limb mobility were excluded from the study. A 10-camera marker-based motion capture system (Qualisys, Gothenburg, Sweden) was used to capture upper extremity kinematic data at 170 Hz. Simultaneously, markerless motion capture data was captured by an 8-camera system (Theia Markerless Inc., Kingston, ON, Canada) at 85 Hz. The participants completed a modified BBT, which involved them moving small wooden blocks as quickly as possible from one compartment over a divider to the other in 60 seconds, aiming for a smaller square indicated on the block release side (Figure 1). They completed three trials for each limb. Shoulder, elbow, and wrist joint angles of the dominant limb were calculated and exported from Visual3D (C-Motion, Inc. Germantown, MD). A custom MATLAB code was used to determine the middle 15 successful block movements within each trial, for each system and calculate overall average joint angles, including shoulder flexion (FLEX), abduction (ABD), and internal/external rotation (IR/ER), elbow FLEX, forearm pronation (PRO), wrist extension/flexion (EXT/FLEX) and radial deviation (RD). The data from both systems were demeaned to account for offset bias and the root mean square deviation (RMSD) was calculated between the two systems. It is hypothesized that the RMSD values would be less than 6°.


Results, Conclusions, and Discussions:: The RMSD values for the demeaned joint angles were all less than 6° (Table 1). The resulting joint kinematics during each successful block movement was similarly captured by the marker-based and the markerless systems. Larger errors existed for shoulder IR/ER and forearm PRO; this is expected due to limited landmark identification in the markerless system to accurately calculate long-axis rotations. The larger errors are also supported by Kanko et al. for angles in the transverse plane.³ It should also be noted that the hypothesized RMSD value was determined based on the lower extremity validation completed by Kanko et al. using markerless motion capture, so further exploration into expected RMSD values for the upper extremity is needed.

Upper extremity joint angles were within 6° of each other when captured using a marker-based and markerless system. The resulting RMSD values indicate the ability to use markerless motion capture systems in non-laboratory settings including in the clinic to examine upper extremity function.




Acknowledgements (Optional): :

References (Optional): : 1. Skurowski P, Sensors, 2022, 22, p. 4076 2. Wade L, PeerJ, 2022, 10 3. Kanko RM, J Biomech, 2021, 127 4. Lahkar BK, Front Sport Act Living, 2022, 4, p. 295