INTRODUCTION
Subject-specific musculoskeletal models show promise in adult spinal deformity prevention and treatment [1–3]. Unfortunately, fully subject-specific models rely on comprehensive 3D imaging that is often unavailable in typical clinical settings. Conversely, generic models are more accessible but potentially less accurate than the subject-specific models. In this context, muscle properties emerge as a potentially important variable influencing the predictive accuracy. This research explores the circumstances under which the generic muscle parameters are adequate, pinpointing situations that necessitate subject-specific modeling. The objective of this study is to identify which muscle parameters, and in which specific simulations, exhibit significant differences when implementing generic versus subject-specific properties.
METHODS
We implemented a musculoskeletal modelling approach (OpenSim 3.3; Figure 1) following established research [4,5] that included generic (average of 250 subjects from the Framingham Heart Study [6,7]) and subject-specific models. Our research focused on four primary muscle parameters: geometry-path (muscle route from origin to insertion), maximum-isometric-force (derived from muscle cross-sectional area), optimal-fiber-length (the length at which a muscle fiber produces its maximum force), and tendon-slack-length. We simulated 11 different static postures, including neutral standing and trunk flexion or extension positions, each chosen based on recommendations of Mokhtarzadeh et al. [7] related to maximum Load-to-Strength ratios. Five subject-specific categories were investigated, with the foremost being a fully subject-specific model. The other four categories evaluated scenarios where one muscle parameter was reverted to generic properties. Variations in axial compression load from L5 to T1 between the fully subject-specific and adjusted models were quantified using the Relative Difference Percentage. A linear mixed-effects model was applied to analyze the influence of muscle parameters on predicted spine compression loads, incorporating covariates such as posture, age, and gender.
RESULTS
Findings showed subject-specific muscle geometry-path and maximum-isometric-force significantly impacted spinal compression loads, with average increases of 13% and 8% respectively (Figure 2). Further, the differences between the subject-specific and generic model predictions were dependant on the posture as shown in Figure 2 (geometry-path p<0.001; max-isometric-force p=0.005). Conversely, parameters such as optimal-fiber-length (p=0.053) and tendon-slack-length (p=0.68) showed minimal effect on load predictions, with only 2% and 1% change, respectively. Notably, high trunk flexion postures were more affected by generic muscle parameters, underlining a posture-specific sensitivity. In contrast, typical standing postures had negligible differences, typically staying below the 5% variation threshold.
DISCUSSION
Our findings underscored the significant impact of individual muscle parameters, particularly the geometry-path and maximum-isometric-force, on the outcomes of biomechanical simulations. The pronounced variations observed in postures characterized by high trunk flexion emphasized the necessity of subject-specific data in such scenarios. Conversely, the minimal deviations observed in standing and low-flexion postures suggested that, when subject-specific muscle data is unavailable, these postures can be reasonably approximated using generic muscle parameters. To translate our findings into clinical application, it is recommended to prioritize standing and low-flexion postures as primary simulations in musculoskeletal modeling for spinal deformity assessments.
Figure 1. Musculoskeletal models in various postures.
Figure 2. Deviation in load prediction between generic and fully subject-specific models, showing average deviations of 13% and 8%.