Interpretable Machine Learning on Simulation-Derived Biomechanical Features for Hamstrings-Quadriceps Imbalance...

Area of Science:

• Biomechanics
• Sports Medicine
• Machine Learning

Background:

• Hamstrings-quadriceps (H-Q) imbalance is a significant biomechanical marker linked to knee instability and increased injury risk in running.
• Accurate assessment of H-Q imbalance is crucial for injury prevention strategies in athletes.

Purpose of the Study:

• To introduce a digital machine learning framework for estimating H-Q imbalance.
• To utilize biomechanical features from synthetic running trials for H-Q imbalance detection.
• To develop a model that conceptually maps to inertial sensor data for practical application.

Main Methods:

• A reduced musculoskeletal model generated 573 synthetic running trials for 160 virtual subjects at three speeds.
• Key biomechanical predictors included dynamic H:Q ratio (H:Qdyn), knee moment limb symmetry index (LSI), and co-contraction index (CCI).
• A gradient-boosting classifier was trained and evaluated using metrics like ROC-AUC, PR-AUC, balanced accuracy, F1, and Brier score.

Main Results:

• The machine learning model achieved high performance across all metrics: ROC-AUC 0.933, balanced accuracy 0.943, PR-AUC 0.918, F1 0.940, and Brier score 0.056.
• Dynamic H:Q ratio and knee moment symmetry were identified as the most influential predictors.
• Co-contraction index provided complementary insights into muscular coordination.

Conclusions:

• Simulation-derived biomechanical data can effectively train machine learning models for assessing muscular balance.
• The developed framework demonstrates interpretable machine learning for transparent H-Q imbalance assessment in sports medicine.
• This approach has the potential to inform injury prevention and performance enhancement in runners.