Segmentation and motion estimation of stent grafts in abdominal aortic aneurysms

A. Klein

Promotor: C. H. Slump and L. J. Schultze Kool
Copromotor: W. K. J. Renema
Universiteit Twente
November 22, 2011

Abstract

Patients with an Abdominal Aortic Aneurysm (AAA) have a high risk of dying due to the rupture of a dilated aorta. Endovascular aneurysm repair (EVAR) is a technique to threat AAA, by which a stent graft prosthesis is implanted in the aorta of the patient, which takes the pressure off the aneurysm wall. Due to its minimal invasive character, this intervention has smaller risks for the patient compared to the conventional approach (in which the unhealthy aorta is replaced with an artificial vessel using open surgery). Yielding good results on the short term, this technique has increased in popularity. Unfortunately, due to effects such as metal fatigue, leakage and stent migration, this technique is less successful on the long term. These problems are caused by the forces applied by the pressure waves of the blood flow. It is therefore important to understand the motion behavior of the stent graft inside the human body. This idea is the main motivation for the research described in this thesis. The technique of ECG-gated CT can be used to obtain multiple 3D images of the patient. Each of these images corresponds to a different phase of the heart cycle. Therefore this technique enables measuring motions inside the patient. We have performed experiments to study the possibilities and limitations for using this technique to measure the motions of interest. We concluded that this technique is indeed suitable to study the motions of stent grafts in AAA. Having obtained the set of 3D images, the goal is to measure the motions of the stent graft. This is done in two steps. The first is segmentation: detecting where the stent is located and subsequently creating a geometric model of the stent. This geometric model is represented as a graph consisting of nodes that are connected by edges. The nodes represent the corners and crossings in the stent’s frame, and the edges represent the wires in between. The second step is registration, by which the deformation between the different images is calculated. Our first attempted segmentation approach was to sample 2D slices (orthogonal to the centerline of the stent) from the 3D images. The points where the metal frame penetrates the slice are then detected. By repeating this process while tracking along the centerline of the stent, a model of the stent can be obtained. Unfortunately, we encountered many problems with the tracking part of this method related to bifurcations and overlapping parts of the stent. To overcome these problems, we decided to work on an approach to segment the stent directly in 3D. First, a set of seed points is detected in the 3D image, based on a few simple criteria. These points are connected to each other using a modified version of the minimum cost path (MCP) method, which is an algorithm that tries to find the optimal route between two or more points. This results in a graph with many edges, which is processed to obtain the final geometric model of the stent. Next is the registration step, in which the deformation fields between the different 3D images are calculated. For this purpose we developed a groupwise registration method which ensures that the transformation fields do not fold. The motion of the stent is determined from the deformation fields and incorporated in the geometric stent model. The obtained dynamic model can be used for visualization of the motion of the stent, as well as for performing further calculations such as estimating the forces in the stent. Clearly, the proposed method is just a small step on the road to better care for patients with AAA. The method should be seen as a tool that enables further quantitative research to motions of stent grafts. These studies will provide new insights in the behavior of the stent graft in vivo. We expect that this will enable the detection and prediction of stent failure in individual patients, and can help in designing better stent grafts in the future.

A pdf file of this publication is available for personal use. Enter your e-mail address in the box below and press the button. You will receive an e-mail message with a link to the pdf file.