Anatomical model generation

		Polyp embedded in the fundus of the uterine cavity. The textures were taken from original endoscopic images. Growing polyp in the fundus. The yellow surface represents the endometrium, the grey particles the myometrium, and the red particles the tumour tissue. Volumetric visualization of growth factor concentration. Simulation result of the vascular growth model.

A crucial component for effective training with surgical simulators is the possibility to provide a different surgical scene in every training session, thus approximating the day-to-day experience of the medical personnel. Different strategies have been investigated for the two main tasks, namely the generation of variable models of the healthy organ (uterus) and creation of different pathologies that can be found therein (polyps and myomas).

Conceptually, the model generation of the healthy organ is based on the integration of an existing set of examples into a database, and the derivation of new organ instances is based on statistical shape analysis. The parameters provided for the derivation of new organ models represent standard clinical measurements used by gynecologists to specify the uterine anatomy, therefore allowing for an intuitive specification of new training scenes.

Three different approaches for the generation of pathologies and variable instances of these pathologies have been investigated, namely a skeleton based design, a cellular automaton and a particle based growth model. From the different tumor growth strategies developed, the growth model based on interacting particles is the most advanced in terms of biological processes represented. The model produces realistic shapes of macroscopic findings and has a high temporal resolution, i.e. the development of the tumor can be observed in relatively small time steps. The model can be extended to incorporate more factors that influence the tumor growth on a macroscopic level, such as the vascularization or biomechanical stresses.

This direction is followed in the second thread of this project. Vascular systems do not simply influence organ appearance as part of their surface texture, but also behave like physical objects with certain mechanical properties. In particular, they will deform along with the hosting tissue and lead to bleeding when cut through. The ultimate goal is to provide a tool which, given a 3D representation of a given tissue/organ and an intuitive set of physiologically meaningful parameters, will generate vascular structures in an arbitrary anatomical region. Such systems are not expected to carry only geometrical information but also provide data on mechanical properties of the vascular system and the related blood flow.

The developed macroscopic flow network model to address the above mentioned demands includes the formation of a primitive capillary plexus prior to maturation of the vascular system and treats its later development as a dynamic growth controlled by biophysical factors. This way the remodeling of the vascular system can be described, and full information on biophysical properties and hemodynamic conditions in the system can be provided at any time. The simulation takes into account basic experimental knowledge of the growing process, namely endothelial cell proliferation and migration, and their modulated response to changes in local growth factor concentrations. The achieved results correspond well to experimental findings.