Prä- und postoperative Segmentierung und virtuelles Stenting von Aneurysmen und Stenosen

Over the past years, medical image processing has become very important in the different treatment stages of somatic diseases. Medical decisions are assisted by computers throughout diagnosis, monitoring, therapy planning, execution and follow-up examinations. In this thesis, contributions to support the computer aided treatment of artery diseases – artery enlargement (aneurysm) and artery contraction (stenosis) – are achieved. If an intervention takes place, there are two different treatment alternatives for this kind of artery diseases: open surgery and minimally invasive (endovascular) treatment. Computer assisted optimization of endovascular treatments is the main focus of this thesis. In the case of an endovascular treatment it is very important to choose an appropriate prosthesis (stent) based on critical patient data as precise and as fast as possible. The choice has to be done before the intervention takes place (preoperative), because it is not possible to change the stent during the intervention without an additional open surgery. Thereafter, regular follow-up examinations are necessary to verify if the stent works properly and to detect holes in the prosthesis (endoleaks) or the movement of the prosthesis (stent migration). For the treatment phases diagnosis, therapy planning and the follow-up examinations of an artery disease, different segmentation methods are developed and presented in this thesis. With these segmentation methods it is possible to monitor aneurysms and stenosis before and after an intervention and therefore support the physician with these time-consuming procedures. The different segmentation methods depend on the Active Contours, the Active Appearance Models and a graph based approach. Among these methods, the graph based approach provided the best results and a prototype has already been realized for clinical evaluation. Therapy planning is supported by a computer-aided simulation of stents (virtual stenting) before an intervention. Currently, a stent is chosen on the basis of CT-scans from a patient. However, with the virtual stenting of this thesis, it is possible to simulate different stents in the patient data from the clinical routine. Thereby, it will be possible to decide whether a chosen stent has proper dimensions and should be used during the following operation. The stent simulation depends on the Active Contours Methods in 3D and is eligible for non-bifurcated and bifurcated stents (Y-Stents). Amongst others, the collision forces that appear in the bifurcation area of Y-stents are considered for a realistic simulation. Furthermore, a virtual stenting approach for the carotis has been developed that pays attention to the behavior of the artery wall during the stent expansion. In addition to the virtual stenting, a catheter path simulation has been developed. The path of the catheter is determined within two steps. The first step calculates an initial path through the artery. For this purpose, Dijkstra’s shortest path algorithm has been modified. The second step uses an Active Contour Model to align the initial path inside the artery. All methods are introduced in detail in this thesis and are evaluated with phantom datasets and real patient data from the clinical routine. Additionally, the clinical prototypes that are based upon these methods are presented.