Development and Cross-Platform Benchmarking of an ArtiSynth Lower-Limb Gait Model Against OpenSim

Background:In biomechanical research, combined simulation of multibody (MB) and finite element (FE) models enable the simultaneous study of global dynamics and tissue-level mechanics. Yet validated pipelines that bridge the MB and FE domains are often complicated and involve multiple software tools. ArtiSynth is an open-source Java-based biomechanical simulation framework for combined MB-FE simulations with forward and inverse simulation capabilities and FE contact modeling. It features an importer for OpenSim models to be used as foundation for combined MB-FE models. However, the performance of OpenSim models in the ArtiSynth environment has not been evaluated to this day. Methods:To address this gap, we developed an ArtiSynth lower-limb model based on an OpenSim model and evaluated it across multiple subjects and trials. Motion and force data from 100 trials of 30 adult healthy participants were processed and used in OpenSim and ArtiSynth inverse/forward computations. A representative trial was selected per participant using the SMaRT algorithm and pooled for Statistical Parametric Mapping (SPM, paired t-tests). Results:Albeit statistically significant differences in most evaluated features, the results showed great kinematic reproducibility (max RMS joint angle error of 2.81∘. Partial kinetic agreement was observed (max RMS joint moment error of 0.49Nm/kg) and substantial differences in muscle recruitment (max RMS muscle force error of 262.48N). Recommended physical plausibility thresholds for forward dynamics were violated approximately 90% of the time in both frameworks, indicating a limitation not unique to ArtiSynth. Conclusions:Although the current results are promising, future research is needed to reduce current residuals and improve muscle recruitment strategies in ArtiSynth before the model can serve as physiologically reliable foundation for combined MB-FE studies.