| Περίληψη: | The development of an accurate subject specific finite element model of the knee anatomy
can be helpful to look into the mechanics of the knee joint. The segmentation of computed
tomography (CT) or magnetic resonance (MR) images is a well-established method to
obtain subject specific geometrical models. Nevertheless, the manual segmentation is a laborious
and time consuming task that heavily relies on the performance of the expert. Multi-atlas
segmentation is becoming one of the most successful automatic image segmentation techniques,
that has been widely tested for brain images but the effort concerning the knee complex is less.
Upon segmentation, the 3D surface is reconstructed using mesh generation and refinement techniques,
necessary for the subsequent biomechanical modelling task. The quality and robustness
of the simulation strongly depends on the contour of the volumetric meshes derived from the
3D surfaces and the shape of the elements. While hexahedral elements are expected to provide
better convergence properties, the generation of geometries composed of hexahedral elements is
a task that requires high expertise and automatic methods do not exist for complex geometries.
To realize a scheme that generates finite element models of the knee complex from MR images,
this thesis utilizes a multi-atlas segmentation method to automatically delineate the regions of
interest in MR images utilizing a set of atlases and an anatomy-adopted meshing algorithm that
automatically creates structured meshes with well shaped hexahedra for the soft tissues of the
knee.
|
|---|
