EP9.1 Agreement and Accuracy of OpenCap-Derived Joint Kinematics for Patients with Prior Slipped Capital Femoral Epiphysis

Abstract Objectives Femoral deformity from slipped capital femoral epiphysis (SCFE) causes movement pattern alterations to prevent painful hip impingement. Understanding these alterations can inform diagnosis and treatment, however, socioeconomic disadvantages can prevent adolescents with SCFE from attending clinic visits. The objective of this study is to evaluate the concurrent validity of OpenCap markerless motion capture, which could allow remote movement assessment, with conventional optoelectronic gait analysis in patients with SCFE. Methods Eight participants (1F/7M, age: 15.3±1.7 years, BMI: 30.2±4.8 kg/m2) with SCFE treated with in-situ pinning 1-3 years ago underwent gait analysis using a 57-marker set, a 12-camera optoelectronic motion capture system (Vicon Motion Systems, Ltd.), and 3 force plates (Advanced Mechanical Technology, Inc.). Subjects performed a minimum of three trials each of three tasks: 1) walking at self-selected speed, 2) double-legged squat, and 3) sit-to-stand-to-sit. Ankle, knee, and hip joint angles were calculated in Visual3D (C-motion Inc.). Concurrently with the marker-based collection, two iOS devices captured simultaneous video recordings for automated biomechanical assessment using the OpenCap (www.opencap.ai) web-based application. Ankle, knee, and hip joint angles were extracted from OpenCap, time-synchronized with the in-lab marker-based data using cross-correlation, and normalized to 0-100% of the given task. Any trials with processing errors were removed, and the remaining successful trials were averaged for each task for each subject. The Pearson correlation coefficient (r) and root mean squared error (RMSE) between the average marker-based and OpenCap data were computed to assess waveform agreement and accuracy, respectively. Results Agreement and consistency were best for hip (median r = 0.982-0.993) and knee (r = 0.976-0.999) flexion, whereas hip internal rotation (r = 0.163-0.932) and ankle inversion (r = 0.239-0.533) demonstrated the lowest agreement and consistency. Errors were smallest for ankle dorsiflexion (RMSE = 7.0°-8.0°) and knee flexion (RMSE = 7.2°-8.0°), whereas hip internal rotation (RMSE = 15.7°-17.5°) demonstrated the highest errors. Conclusion Agreement and accuracy of OpenCap-derived kinematics appear to be subject-, task-, and joint-specific. Further algorithm refinement to ensure kinematic errors of <5° is crucial for realistic clinical implementation of OpenCap for patients with altered musculoskeletal biomechanics.