Concurrent Validity of the Human Pose Estimation Model “MediaPipe Pose” and the XSENS Inertial Measuring System for Knee Flexion and Extension Analysis During Hurling Sport Motion

Hurling is a stick-and-ball field game, predominantly played in Ireland. Knee injuries are frequent in Hurling and are associated with ongoing issues such as reinjury and decreased performance. Analysis of sport motion is crucial for understanding the quality of the technical aspects of sport skill, and to identify risk factors of injury. The aim of this study was to compare the use of a human pose estimation (HPE) model “Mediapipe Pose” to the XSENS MVN Link inertial measuring unit (IMU) system for calculation of knee flexion and extension during Hurling sport motion. To assess concurrent validity, 10 trials of an overhead catch were performed by an experienced Hurling player where knee flexion and extension angles of the right knee were captured simultaneously by an inertial measuring unit (IMU) system and a video camera. A HPE model “MediaPipe Pose” was subsequently applied to the video footage. The correlation between the knee joint angles measured by the two systems was assessed using Spearman's rho correlation coefficient. Results suggest a significant positive correlation between the IMU system and the HPE measurements for the right knee $(\rho=\ 0.838^{**},\ \mathrm{p} < 0.01)$ . This indicates good concurrent validity of the HPE model for calculating and tracking knee flexion and extension angles. However, the mean absolute error in knee joint angles between the IMU system and HPE was between 4.91 and 12.2°. These results provide valuable insights into the use of HPE and computer vision techniques as a valid and non-obtrusive method of calculating joint angles in Hurling sport motion. The authors aim to conduct an additional validation test to verify these results. Future research shall test a greater sample of subjects and enhance the analysis methods by incorporating high quality depth cameras. This information may be used by biomechanists and sport scientists to explore the potential for applying artificial intelligence and biomechanical analysis to improve performance in other sports or applications.