A Multibody Model for Predicting Spatial Distribution of Human Brain Deformation Following Impact Loading

Area of Science:

• Biomechanics
• Neuroscience
• Computational Modeling

Background:

• Concussive brain injuries have long-term consequences, necessitating accurate predictive tools for mechanical responses to impact.
• Current finite element models (FEM) provide estimates but lack the speed for real-time analysis of brain response to dynamic loading.

Purpose of the Study:

• To develop a rapid computational model for predicting the brain's mechanical response to rotational accelerations.
• To enhance prediction accuracy by integrating a multibody model with machine learning techniques.

Main Methods:

• Developed a multibody spring-mass-damper model (MBM) to estimate regional brain motion under rotational acceleration.
• Employed machine learning (ML) to combine MBM predictions with loading kinematics (rotational acceleration, velocity).
• Validated the hybrid MBM-ML model against FEM predictions using independent datasets, including sports injury data.

Main Results:

• The MBM alone showed correlation but not precise prediction of FEM response (18.4% average relative error).
• The hybrid MBM-ML model significantly reduced prediction error compared to MBM alone (9.8% average relative error).
• The hybrid model demonstrated good correlation with FEM on an independent sports injury dataset (16.4% average relative error).

Conclusions:

• The hybrid MBM-ML approach provides a rapid and reasonably accurate method for predicting brain mechanical responses to various impact types.
• This new tool can quickly assess the consequences of impact loading across different brain locations, aiding in injury assessment.