Hypersensitivity of trunk biomechanical model predictions to errors in image-based kinematics when using fully...

Recent advances in medical imaging techniques have allowed pure displacement-control trunk models to estimate spinal loads with no need to calculate muscle forces. Sensitivity of these models to the errors in post-imaging evaluation of displacements (reported to be ∼0.4-0.9° and 0.2-0.3 mm in vertebral displacements) has not yet been investigated. A Monte Carlo analysis was therefore used to assess the sensitivity of results in both musculoskeletal (MS) and passive finite element (FE) spine models to errors in measured displacements. Six static activities in upright standing, flexed, and extended postures were initially simulated using a force-control hybrid MS-FE model. Computed vertebral displacements were subsequently used to drive two distinct fully displacement-control MS and FE models. Effects of alterations in the reference vertebral displacements (at 3 error levels with SD (standard deviation) = 0.1, 0.2, and 0.3 mm in input translations together with, respectively, 0.2, 0.4, and 0.6° in input rotations) were investigated on the model predictions. Results indicated that outputs of both models had substantial task-dependent sensitivities to errors in the measured vertebral translations. For instance, L5-S1 intradiscal pressures (IDPs) were considerably affected (SD values reaching 1.05 MPa) and axial compression and shear forces even reversed directions as translation errors increased to 0.3 mm. Outputs were however generally much less sensitive to errors in measured vertebral rotations. Accounting for the accuracies in image-based kinematics measurements, therefore, it is concluded that the current measured vertebral translation errors at and beyond 0.1 mm are too large to drive biomechanical models of the spine.