Data-driven uncertainty quantification in computational human head models

Computational models of the human head are promising tools for estimating impact-induced response brain, and thus play an important role in prediction traumatic brain injury. Modern biofidelic model simulations associated with very high computational cost, high-dimensional inputs outputs, which limits applicability traditional uncertainty quantification (UQ) methods on these systems. In this study, a two-stage, data-driven manifold learning-based framework is proposed UQ models. This demonstrated 2D subject-specific model, where goal to quantify simulated strain fields (i.e., output), given variability material properties different substructures input). first stage, method based multi-dimensional Gaussian kernel-density estimation diffusion maps used generate realizations input random vector directly from available data. small number provide input-output pairs training surrogate second stage. The employ nonlinear dimensionality reduction using Grassmannian maps, process regression create low-cost mapping between reduced solution space, geometric harmonics space Grassmann manifold. It that highly accurate approximations while significantly reducing cost. Monte Carlo propagation. highlight significant spatial variation uncertainty, reveal key differences among commonly strain-based injury predictor variables.