Reliability and validity of knee valgus angle calculation at single-leg drop landing by posture estimation using...

Area of Science:

• Biomechanics
• Sports Medicine
• Artificial Intelligence

Background:

• Dynamic knee valgus (DKV) is a key factor in anterior cruciate ligament (ACL) injury prevention.
• Traditional 3D motion analysis systems for DKV evaluation are costly and inaccessible for widespread athlete screening.
• Markerless motion capture and pose estimation offer potential alternatives, but their reliability and validity for DKV assessment are unproven.

Purpose of the Study:

• To evaluate the reliability and validity of dynamic knee valgus (DKV) calculations using AI-powered pose estimation.
• To compare DKV measurements from pose estimation with gold-standard 3D motion analysis systems.
• To determine the optimal method for DKV calculation using pose estimation for ACL injury prevention.

Main Methods:

• Fifteen participants performed single-leg jump landings.
• Knee joint angles were calculated using AI pose estimation (MediaPipe Pose) and 3D motion capture (VICON MX).
• Calculations included absolute values and changes from initial ground contact (IC); reliability and validity were assessed using ICC and Pearson's correlation.

Main Results:

• AI pose estimation showed significantly higher DKV values than 3D motion analysis (error range 18.83-19.68°).
• Absolute DKV values lacked significant concurrent validity, while the change from IC demonstrated good reliability and validity.
• Inter-rater reliability for absolute DKV was low, but good for the change from IC.

Conclusions:

• AI-powered pose estimation is a practical tool for DKV assessment in ACL injury prevention.
• Normalizing DKV calculations by the angle at initial ground contact is crucial for reliable and valid measurements.
• This approach provides a more accessible method for athlete screening and injury prevention strategies.