Detail

Rolf M. Koch

PhD Thesis 13912
ETH Zürich
Institute for Scientific Computing
12/2000.

Keywords:

Pages: 176
Language: English

Methods for Physics Based Facial Surgery Simulation

This thesis presents a framework for facial surgery simulation, the prediction of the postsurgical outcome, and facial expression synthesis which is constructed on surface based and volumetric finite element modeling. There is a wide spectrum of facial malformations and diseases which maxillofacial and cranio-facial surgeons have to treat. These include, but are not restricted to, injuries and tumors as well as deformities due to inherited syndromes or developmental disorders. Furthermore, facial surgery has to strive for the reconstruction of a balanced face, as even very fine variations in facial proportions can affect the appearance of a face strongly and thus distort its harmony. Therefore, both surgeon and patient have a strong need for a method which enables them to compute highly realistic pictures of the expected post-surgical shape during the planning of a surgical procedure. In addition to the prediction of the postsurgical appearance, it would be very helpful not only to predict the static face, but also the dynamic behavior and emotional expressions. Besides surgery simulation, a physics based facial model could also be used in the field of lifelike anima-tion. In addition, the development of lifelike avatars for applications in virtual environments, for instance conference systems, is of great importance in recent computer graphics research. More generally, any application that could benefit from a humanlike interface could also make good use of a facial model. For these purposes, two facial models have been developed differing in regard to the exact anatomical representation used. These models are based on either a combination of an elastic surface and a spring mesh, or on elastic volume elements that are simulated using the fundamental theory of linear elastomechanics and the finite element method. Especially in facial animation, facial models are mostly adapted from a template face instead of generated directly from the underlying data set. Furthermore, unsatisfactory approximations of the patient�s skull surface have been used. Since the skull and the exact representation of the facial surface are essential for surgical applications, the facial models presented in this thesis are constructed based on real patient volume scan and laser range scan data. Furthermore a tissue segmentation based on CT data is applied to enable the discrimination of different tissue types and to take into account the variation of tissue stiffnesses as well as tissue incompressibilities throughout the body. A principal contribution made in this thesis is the combination of the physical correctness of volumetric or surface-based finite element simulation with the superior quality of a C1-continuous surface representation to optimize both accuracy and rendering quality. Another important contribution is a thorough evaluation of the volumetric approach with a group of test patients. With this aim, a prototype application was built which is based on data available from individual patients in a clinical environment. This includes CT scans of the pre- and postsurgical situation in combination with high resolution laser range surface scans. Both individual parameters, such as varying stiffness and elasticity of tissue, and the bone movements introduced by the surgeon are taken into account in order to simulate as accurately as possible the outcome of real surgery. With validation in mind, the surgical procedures are re-simulated on the test group of patients with cranio-maxillofacial abnormalities.