Improvements in finite element models of spinal cord injury and interpretation of mechanical outputs using machine learning

This thesis explores the correlation between mechanical impact and neurological damage in spinal cord injury (SCI) through computational modeling and machine learning (ML). It emphasizes on the role of the cerebrospinal fluid (CSF) boundary conditions and morphology in SCI models and the classification of injured elements using ML algorithms. Findings reveal the major influence of CSF boundary conditions and morphology on the predicted mechanical outcomes, highlighting the importance of proper modelling choices for enhancing the models' biofidelity. Furthermore, the integration of supervised ML techniques facilitates the classification of injured elements based on FE model results, providing valuable insights into specific damage mechanisms. Additionally, the feature importance analysis and injury threshold estimation reveal distinct susceptibility patterns in gray and white matter tissues. This work contributes to the development of more reliable computational models and innovative approaches for SCI prevention and treatment, addressing a critical need in the field of SCI research.