Reliability of Precision in Motion Software

Background: There is a current lack of affordable 3-D motion analysis systems available to physical therapists for documenting objective data. Objective: Compare data from inexpensive 3-D motion capture system using the Kinect, to the Codamotion, a traditional active marker, motion analysis system. Determine data reliability. Design: Reliability-method comparison study Methods: Seventeen subjects (9 females, 8 males) were instructed to walk down marked runway, arms across chest while both systems recorded data. Data was collected with Kinect on the same and opposite side as the Coda. Kinect and Coda data were interpolated and normalized for comparison. A program for heel strike detection was written in R for both systems, which determined stride length and stride time. Interclass correlation coefficients (ICC 3,1) were calculated for stride length, stride time, and joint angles for all subjects. Results: All stride length ICC ≥ 0.70 shows acceptable clinical reliability. One of four stride time ICC \u3c 0.70 for opposite leg step 1 likely due to occlusion of test leg by other leg during step. Joint value comparisons show high variability in reliability of Kinect software. Seventy-seven % of the same side knee data and fifteen % of the hip same side data from Kinect is clinically reliable. Limitations: Kinect and Coda use different skeletal models. Conclusions: Kinect software updates are needed to improved accuracy. Kinect system was significantly easier to use than Coda and may eventually be a cheaper alternative to traditional motion analysis systems that could be used by clinicians to monitor patient progress