Patient-Specific Adaptation of a Mechano-Regulatory Bone-Healing Model Using Longitudinal Loading Data

Patient-specific simulation frameworks are increasingly used to support orthopaedic trauma decision-making, yet robust coupling of longitudinal loading data to mechano-regulated healing models remains challenging. Here, we present a workflow that links patient-specific finite element mechanics to an established mechano-regulatory reaction-diffusion healing model and updates boundary conditions longitudinally using weekly knee joint forces obtained from musculoskeletal simulation. Rather than proposing a new mechanobiological regulation law, the study addresses three methodological questions: how strongly stimulus choice affects predicted tissue differentiation on identical patient-specific finite element models, whether an Isaksson-based healing model can be stabilised for robust long-term implicit simulation under repeated load updates, and whether computationally efficient linear-elastic mechanics is sufficient for repeated longitudinal simulations under the present loading regime. Mechanical stimulus formulations from the literature were benchmarked on identical patient-specific finite element models to quantify their impact on predicted tissue differentiation. To enable stable long-term implicit simulations, we introduced a numerically robust modification of the available-space saturation terms in the proliferation and matrix production kinetics. In a single distal tibia clinical case, the simulation predicted predominantly bone-permissive stimulus conditions, progressive bone-matrix accumulation initiating from existing bone boundaries, and a pronounced reduction of implant-to-callus stress ratio over the first 40-50 days, indicating early critical fixation demand. Synthetic radiographs generated from the final Young's modulus distribution showed qualitative agreement with the dominant bridging pattern and regions of reduced mineralization observed in follow-up X-rays, whilst suggesting over-persistent callus consistent with the absence of secondary remodelling. Sensitivity analyses showed that small-strain linear elasticity provides near-equivalent stimulus-class distributions at substantially lower computational cost than nonlinear alternatives, supporting its use for high-throughput longitudinal pipelines. These results illustrate feasibility in a single clinical case; cohort-level validation and predictive performance for delayed union/non-union are topics of ongoing and future multi-case studies.