Comparing novel smartphone pose estimation frameworks with the Kinect V2 for knee tracking during athletic stress...

Area of Science:

• Biomechanics
• Sports Medicine
• Computer Vision

Background:

• Anterior cruciate ligament (ACL) injuries are common in athletes.
• Accurate assessment of knee kinematics is crucial for injury prevention.
• Current methods for kinematic analysis can be cumbersome and expensive.

Purpose of the Study:

• To evaluate the accuracy of Google's MediaPipe framework for knee kinematics.
• To compare MediaPipe's performance against the Microsoft Kinect V2.
• To explore the potential for smartphone-based ACL injury risk assessment.

Main Methods:

• 254 varsity athletes were analyzed using Kinect V2 and a smartphone app with MediaPipe.
• Athletes performed three athletic stress tests at a 2.5m distance.
• Knee angles were extracted and compared between the two systems, with Kinect V2 as ground truth.

Main Results:

• Small differences in knee angles were observed in the coronal plane.
• Moderate differences were found in the sagittal plane.
• MediaPipe tended to underestimate knee valgus and sagittal angles compared to Kinect V2.

Conclusions:

• Google's MediaPipe framework demonstrates potential for lower limb kinematics analysis.
• Smartphone-based applications using MediaPipe could enable widespread, low-cost ACL injury prevention.
• Further validation is needed, but results suggest a viable alternative to traditional motion capture.