A comparison between manual and automated event detection for a drop vertical jump task using motion capture

Background: The use of movement screens as a clinical tool for injury risk assessment requires variables to be extracted across specific phases of interest. While manually selecting task events is the traditional method, automated event detection is an effective technique that maintains consistency across a cohort. This study aimed to examine variations in event identification, comparing manual detection and the application of an automated algorithm, with a specific focus on a drop vertical jump task. Methods: Thirty participants cleared to return-to-play after anterior cruciate ligament reconstruction and thirty controls were tested. For the automated event detection, normalized vertical ground reaction force and the velocity of the sacrum marker were used to identify five events during the drop vertical jump: initial contact, end of loading, end of propulsion, second contact, and end of second loading. Two raters manually selected events and were compared to the event times of the automated algorithm. Findings: Manual event detection exhibited excellent reliability Significant differences between manual and automated detection were observed, particularly at events indicating the lowest squat position (Event2 and Event5). Participants who had undergone anterior cruciate ligament reconstruction demonstrated larger differences than controls at Event5, correlating with significant squat depth disparities. Interpretation: While manual event detection demonstrated reliability, automated algorithms revealed differences, specifically in events of the drop vertical jump involving the lowest squat position. The automated algorithm presents potential benefits in reducing processing time and enhancing accuracy for event identification, offering valuable insights for motion capture applications in clinical settings.