Efficient and validated finite element analysis for quantitative assessment of fracture healing

Accurate, objective assessment of fracture healing remains a critical unmet clinical need, as current evaluation relies on subjective radiographic and clinical observations. Finite element (FE) modeling has demonstrated promise for quantifying mechanical competence of healing bone, yet its clinical adoption is hindered by technical complexity and computational cost. This study validates a stepwise voxel-based FE modeling framework designed to enable efficient, automated, and scalable assessment of fracture healing from CT images. Using an ovine tibial osteotomy model with longitudinal CT scans and continuous in vivo strain sensor data as ground truth, we systematically transitioned from a high-fidelity FE reference model to a simplified voxel-based implementation in ParOSol. At each of five levels, single modifications were introduced to geometry, mesh type, boundary conditions, or solver, enabling direct comparison of diagnostic accuracy and computational performance. Across levels, virtual torsional rigidity (VTR) closely tracked healing progression and strongly correlated with the reference FE models (R² ≥ 0.969). When compared with sensor-derived in vivo measurements, the voxel-based model achieved performance comparable to the high-fidelity reference model (R² = 0.769 and 0.788, respectively). Healing times were consistent across models, with no significant pairwise differences, while voxel-based simulations achieved nearly a 3-fold reduction in runtime. Notably, all levels correctly identified delayed and nonunion cases, underscoring diagnostic robustness. These findings demonstrate that voxel-based FE modeling retains diagnostic sensitivity while greatly enhancing computational efficiency. With further automation, this framework could enable CT-driven, patient-specific, and clinically compatible monitoring of fracture healing.