465 - Reliability and Accuracy of a Digital Video Analysis System for Trunk...

#465 Reliability and Accuracy of a Digital Video Analysis System for Trunk and Lower Extremity Kinematics

Sports and Spine

Poster Presented by: J. Paul


J. Paul (1)
A. Petrizzo (1)
A. Patel (2)
E. Godwin (2)
K. Bianco (1)
S. Maier (1)
T.J. Errico (1)
V. Lafage (1)
C. Paulino (2)

(1) NYU Hospital for Joint Diseases, Orthopaedic Surgery, New York, NY, United States
(2) SUNY Downstate Medical Center, Orthopaedic Surgery, Brooklyn, NY, United States


Introduction: Dynamic measures of gait and motion performance have the potential to help surgeons determine what operative and radiologic parameters optimize functional after surgery. For a more complete understanding of the effects of fusion on motion, studies of performance before and after surgery are required, but the current gold-standard 3D motion capture involves expensive, stationary equipment limiting access. This study investigates the potential for an alternative low-cost, portable, 2D automated digital video motion analysis tool.

Methods: To evaluate reliability of 2D video collection and marker-tracking, ten able-bodied controls performed two sessions of gait and range of motion exercises. Gait and functional range of motion exercises were assessed for inter-trial, inter-session, and inter-rater reliability. To compare 2D with 3D analysis, sixteen subjects with a diagnosis of AIS performed formal gait analysis with simultaneous acquisition of two- and three-dimensional data. Dartfish and Vicon generated data sets including the y(longitudinal) and z(vertical) marker position coordinates over an approximately 1m path for 1.0sec during gait. The video generated positions were averaged for distance from the gold standard 3D position and the array of positions were compared for correlation and paired statistical t-test.

Results: Reliability of simple forward and lateral bending was considered good to excellent as the inter-trial ICC of the hip flexion angle was 0.959 and the knee angle was 0.934 on forward bend. The shoulder-pelvis (spine inclination) angle during lateral bending was 0.97 (representative plot shown in Figure). Estimation of rotation angle resulted in ICC of 0.846. Stride length and knee angle were measured during gait on a treadmill resulting in ICC of 0.97 and 0.91, respectively. Hip and knee angles were also measured for the complex task of rising from seated position with ICC of 0.93 and 0.92, respectively. For accuracy of the overground gait, significant positive pairwise correlations (p< 0.05) were obtained when comparing 2D and 3D positions for 4 extremity markers. For all markers, the overall average distance between Dartfish and Vicon generated positions was 5 +/- 3.2cm. Accuracy was greater for fast-moving distal ankle and wrist markers than for slower-moving proximal elbow and knee. Separately, the differences in the horizontal and vertical planes were 5 +/- 4.1cm and 1.4 +/- 1.5cm, respectively.

Discussion and Conclusion: The accuracy and reliability of 2D motion tracking analysis is acceptable for measurement of simple functional range of motion and gait kinematics. The horizontal component of the position contributed most to the error, which likely represents parallax at the beginning and end of the recorded gait. If so, accuracy should improve with stationary movement, such as on a treadmill. Limited motion analysis data exist in spine surgery and are difficult to compare across studies. Taken together, the accuracy, affordability, portability and ease of use may offer a means toward higher-powered future studies of spine and pelvic range of motion.

Figure 1