Artificial Intelligence-Based Exploration of the Relationship Between Running Motion Patterns and Achilles Tendon Stress in Inactive Adults Initiating a Running Program.

The XGBoost model outperformed both Random Forest and Support Vector Regression models in predicting Achilles tendon stress from running biomechanics.

• Three machine learning models were compared: Extreme Gradient Boosting (XGBoost), Random Forest (RF), and Support Vector Regression (SVR).
• All models were integrated with the SHAP (Shapley Additive Explanations) framework for interpretability.
• XGBoost demonstrated superior prediction accuracy among the three models tested.
• The study included 189 inactive adult participants whose running biomechanics were comprehensively assessed.

Ankle dorsiflexion angle below 10.5° was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified ankle dorsiflexion angle as a key biomechanical predictor of Achilles tendon stress.
• The critical threshold was identified at less than 10.5° of ankle dorsiflexion during running.
• This finding suggests that moderately increasing ankle dorsiflexion may be important for decreasing Achilles tendon stress in inactive adults.

Ankle plantarflexion moment exceeding 1.5 N·m/kg was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified ankle plantarflexion moment as a significant predictor of Achilles tendon stress.
• The critical threshold was identified at exceeding 1.5 N·m/kg.
• The authors recommend reducing ankle plantarflexor activation as a strategy for decreasing Achilles tendon stress in inactive adults beginning to run.

Ankle eversion moment exceeding 0.1 N·m/kg was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified ankle eversion moment as a key biomechanical predictor.
• The critical threshold was identified at exceeding 0.1 N·m/kg.
• This threshold was notably low, suggesting even small eversion moments may meaningfully influence Achilles tendon stress.

Hip internal rotation angle exceeding 8.2° was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified hip internal rotation angle as a significant proximal joint predictor of Achilles tendon stress.
• The critical threshold was identified at exceeding 8.2° of hip internal rotation.
• This finding highlights the role of proximal joint movement patterns in influencing distal Achilles tendon loading.

Ankle external rotation angle below 25.3° was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified ankle external rotation angle as a key predictor of Achilles tendon stress.
• The critical threshold was identified at less than 25.3° of ankle external rotation.
• The authors recommend moderately increasing ankle external rotation as a strategy to decrease Achilles tendon stress.

Knee flexion angle below 23.3° was associated with significantly increased Achilles tendon stress.

• SHAP analysis identified knee flexion angle as a significant predictor of Achilles tendon stress.
• The critical threshold was identified at less than 23.3° of knee flexion during running.
• This finding indicates that optimizing proximal joint movement patterns, including knee flexion, may be crucial for decreasing Achilles tendon stress.

Achilles tendon stress was estimated using a combination of OpenSim musculoskeletal modeling and ultrasound imaging in 189 inactive adult participants.

• A total of 189 inactive adults were recruited for the study.
• Running biomechanics were comprehensively assessed for all participants.
• Achilles tendon stress estimation combined OpenSim musculoskeletal modeling with ultrasound imaging.
• The SHAP framework was used to interpret the relationship between running biomechanics variables and Achilles tendon stress.