Do we need medical imaging-informed musculoskeletal models for simulations in healthy adults? A new workflow based on magnetic resonance imaging highlights the importance of personalized geometry.

Musculoskeletal simulations often rely on generic models that may fail to accurately represent individual anatomy. While personalization using medical imaging can enhance model accuracy, it is often assumed to be more critical for pathological cases or pediatric populations, as generic models are typically based on healthy adults. However, even in healthy adults, generic models may not capture individual anatomical variability. In this study, we present a semi-automatic workflow for creating personalized musculoskeletal models based on magnetic resonance imaging (MRI). Our workflow concentrates on creating subject-specific joint centers and muscle paths. It also reconstructs bone surfaces without requiring MRI segmentation. It integrates 3D Slicer and Python scripts, and uses Thin-Plate Spline ('TPS') mapping of anatomically equivalent ('homologous') landmarks from generic models onto participants' anatomy. We applied this workflow to eight healthy participants, generating both generic-scaled and MRI-based models. Simulations were performed using participants' 3D motion capture data. Two model types were compared using a number of parameters, including model geometry, joint kinematics, dynamics, and resultant joint contact forces during one gait cycle. The results revealed clear geometric differences between the model types, with MRI-based models exhibiting a wider pelvis (mean distance between ischial bones was 98.0 ± 5.0 mm in generic-scaled and 11.0 ± 8.0 mm in MRI-based models) and distinct femur/tibia proportions (the mean ratio was 0.92 ± 0.040 in generic-scaled and 0.99 ± 0.033 in MRI-based models). MRI-based models captured systematic anatomical differences between males and females that were absent in generic-scaled models. These geometric differences substantially affected joint loading estimates. MRI-based models consistently produced higher peak joint contact forces with greater inter-individual variation, particularly at the knee joints. Early stance knee peak joint contact force was higher in the MRI-based compared with the generic-scaled model by 0.84 ± 1.28 body weights on average. Despite these differences in geometry and loading, joint kinematics were similar within individuals (mean difference was 0.8 ± 2.47°) and muscle moment arms aligned well with published cadaver data, supporting the validity of the personalization approach. This workflow simplifies the creation of MRI-based musculoskeletal models and challenges the assumption that personalization is unnecessary for healthy adults. The findings reveal significant sensitivity of joint contact forces to individual morphology, emphasizing the importance of personalized models even in healthy populations for biomechanical analyses.