About me
Comparison of artificial intelligence based markerless motion capture with marker based motion capture for single leg hop analysis
Location: 19
Mentor: Dr. Moataz Eltoukhy
Motion capture is used in biomechanics to assess movement patterns in clinical and performance settings. Marker-based systems like Qualisys are considered valid due to their use of physical markers. However, they require extensive setup, including placement, calibration, and correction. Markerless systems such as KinaTrax offer a time-efficient alternative by using computer vision and AI to track movement without physical markers. PURPOSE: Current concerns are whether the data quality of markerless systems align with marker-based systems. This study compares markerless and marker-based motion capture systems during the single leg hop for distance (SLHD) using several key kinematics (e.g. pelvis, trunk, knee). METHODS: Data were collected on 17 ACLR participants (27.12 ± 9.43 years, 1.71 ± 0.08 meters, 70.91 ± 13.13 kilograms, 43.12 ± 13.41 weeks since surgery) from a University of Miami Sports Medicine cohort. Motion capture systems were calibrated prior to data collection and then each participant was fitted with 35 retroreflective markers to be worn while performing SLHD trials. A successful trial required taking off and landing on the same foot with at least one second of stabilization. Participants performed three successful trials per limb with both motion capture systems simultaneously capturing the SLHD. Motion capture data were processed separately on their respective platforms before kinematic analysis in a standard biomechanical modeling software. Trials were normalized from 25 frames prior to SLHD landing through 75 frames following SLHD landing. RESULTS: A total of 95 (R/L: 46/49) trials were included in this analysis to assess agreement in joint angles between motion capture systems using the mean absolute error (MAE). The average (+ standard error) MAE across key kinematics was 19.52 + 0.76 degrees [18.02, 21.02] at the pelvis, 3.84 + 0.34 degrees [3.17, 4.51] at the trunk, 12.49 + 0.58 degrees [11.34, 13.65] at the right knee, and 12.13 + 0.60 degrees [10.92, 13.34] at the left knee. CONCLUSION: Low MAE in the trunk indicates strong agreement between systems for assessments of trunk control during the SLHD while moderate agreement exists between motion capture systems in both knees and the pelvis segment. Future analyses should aim to establish suitable ranges for MAE in healthy controls along with individuals returning from ACLR.
