Predictive Behavior of a Computational Foot/Ankle Model through Artificial Neural Networks

Area of Science:

• Biomechanics
• Computational modeling
• Human joint analysis

Background:

• Computational models are essential for studying human joint biomechanics.
• Model accuracy relies on detailed anatomy and soft tissue properties.
• Literature data often provides approximate tissue properties for these models.

Purpose of the Study:

• To enhance the predictive performance of a computational foot/ankle model.
• To optimize ligament stiffness inputs using artificial intelligence.
• To compare feedforward and radial basis function neural networks for this optimization.

Main Methods:

• Developed a computational foot/ankle biomechanics model.
• Employed feedforward and radial basis function neural networks.
• Optimized ligament stiffness parameters using these neural network approaches.

Main Results:

• Both neural network types provided reasonable ligament stiffness predictions.
• Feedforward neural networks showed better performance based on mean square error.
• Optimized inputs were suitable for implementation into the computational model.

Conclusions:

• Neural network optimization significantly improves ligament stiffness inputs for biomechanical models.
• Feedforward networks offer superior accuracy in predicting these parameters.
• Enhanced models provide more reliable simulations of foot/ankle joint biomechanics.