Abstract: For those studying the biomechanical response of the lumbar spine, anatomical meshes obtained from medical imaging data is quite important. However, such models are generally fixed and can only represent a single subject’s geometry. The objective of this study was to improve our previous lumbar vertebral CAD model such that the parameters are now extracted from the CT scan using a semi-automatic procedure. To illustrate the procedure, first, the transverse cross-sections of vertebral bodies were obtained from an individual at three levels, superior, middle and inferior. Parametric contour curves were fitted onto the vertebral body boundaries using an optimization procedure and the fitting errors are reported here. Five lumbar vertebral bodies were then formed using the lofting operation. Hausdorff distance between the CAD and segmentation models was used to assess the accuracy of the resulting models. The means and standard deviations of Hausdorff distances are also reported here. The adopted optimization process was observed to be resulting coefficient of determination values as high as 0.978. Our new model is expected to lead to dramatic reductions in both time and effort required to build a patient-specific biomechanical model.
Keywords: Parametric model Segmentation Patient specific Lumbar spine Vertebral parameters Spinal curve Medical image processing
