A comparison between automatically generated linear and parabolic tetrahedra when used to mesh a human femur.

Finite element models of bone segments generated from computed tomography data using automatic mesh generation algorithms are becoming common not only in research but also in clinical applications such as computer aided orthopaedic surgery. Especially in the case of the latter application, the models cannot be verified against an experimental measurement, therefore their inherent accuracy should be well known before drawing conclusions based on the calculated results. This study was carried out to assess the performance of tetrahedral solid finite elements with linear and quadratic displacement functions when they are used to mesh the human femur in conjunction with automatic mesh generator methods. Ten-node quadratic tetrahedra (T10) having parabolic displacement functions were compared with four-node linear tetrahedron elements (T4) on the basis of accuracy and central processing unit (CPU) time. From the analyses of 11 finite element meshes, it was concluded that linear tetrahedral elements should be avoided and quadratic tetrahedral elements ought to be chosen for the purposes of finite element analysis of the human femur. When incremental loading and iterative solution is necessary, the coarsest possible T10 mesh compatible with accuracy is needed to minimize computer capacity and CPU time.