Assessing the reproducibility of a subject-specific finite element modelling pipeline for the human metastatic vertebrae

Computed tomography (CT)-based subject-specific finite element (SS-FE) models offer a quantitative approach to assessing the stability of metastatic vertebrae. A critical step in developing SS-FE models is image segmentation, however automatic techniques fail to accurately capture the altered contours of metastatic vertebrae, making manual image segmentation the preferred approach. This introduces operator-dependent steps into the CT-to-FE modelling pipeline that may influence the model's mechanical properties. To assess these dependencies, this study evaluated intra- and inter-operator reproducibility of a CT-to-FE pipeline applied to six lumbar vertebrae, three radiologically healthy and three lytic metastases. All models were developed using a standardised pipeline incorporating densitometric calibration, 1 mm quadratic tetrahedral meshing, and heterogeneous isotropic elasto-plastic material properties, with vertebrae loaded in uniaxial compression to 1.9% apparent strain. Geometrical/volumetrical metrics from segmented images and mechanical metrics from the SS-FE models were assessed. Intra-operator reproducibility was high across geometric and mechanical metrics, whereas inter-operator reproducibility showed modest variability, particularly in local mechanical metrics of metastatic vertebrae. Consistent volumetric inconsistencies between segmentations significantly correlated with variation in predicted failure load of SS-FE models. Overall reproducibility was high, demonstrating the reliability of the CT-based SS-FE modelling pipeline for estimating vertebral strength.