Dissertation Defense of Andreas Keil
: Dynamic Variational Level Sets for Cardiac 4D Reconstruction
An increasing number of cardio-vascular procedures combined with the trend toward less invasiveness of these interventions demand intra-interventional imaging using flexible C-arm systems. Although high speed, dual source, 4D CT angiography recently emerged as a diagnostic imaging modality with a high negative predictive value, the ultimate goal is a 3D visualization of the intra-operative situation, combining diagnosis, navigation, and quantitative evaluation.
The reconstruction of the coronary vasculature from angiographic C-arm sequences, however, is an order of magnitude more complex than that of other anatomy due to the inevitable motion blur associated with these acquisitions and the small size of the structures of interest. A successful and dose-efficient reconstruction algorithm therefore has to compute the cardiac motion in addition to the structural information.
Based on that application, this thesis presents a novel method for purely image-based 4D shape reconstruction without prior data. Adapting and extending mathematical models that originated from other disciplines, new methods for the reconstruction of dynamic shapes from arbitrary projections are developed. The main contributions are the introduction of dynamic level sets, combining implicit shapes with explicit motion information, and the probabilistic derivation of a new kind of energy functionals for symbolic shape reconstruction. Experiments ranging from synthetic over phantom to real data sets prove the feasibility and versatility of the approach.
: Prof. Dr. Daniel Cremers, TUM
: Prof. Dr. Nassir Navab, TUM
: Prof. Dr. Rebecca Fahrig, Stanford University
: Thursday, September 16, 2010
: 16:30 (35min public presentation + approx. 30-40min non-public examination)
: Math/Informatics building (FMI), seminar room MI 03.13.010
, top floor, aisle 13
Technische Universität München
Fakultät für Informatik
85748 Garching b. München