Thu, 23 July, 2009, 14:00-15:00, C1
We will present a new method for the analysis of Magnetic Resonance
(MR) cardiac images with the goal of reconstructing the motion
(deformation) of the right and left ventricular walls. The
reconstructed motion is used to compute key cardiac function
indicators like ejection fraction and regional wall thickening.
The method consists of two main components. The first component is an
intensity-based elastic image registration algorithm that is
accelerated by a non-uniform discretization for the deformation field
and a parallel geometric multigrid solver for the first-order
optimality conditions of the registration functional. In the second
component, given a deformation field, we solve a
biophysically-constrained 4D inverse problem to recover the forces in
the myocardium. The goal of the second step is twofold: (1) extract
features (forces along the myocardial fibers) that effectively
summarize the motion of the myocardium; and (2) inform and improve
existing biomechanical models of the heart. Our main hypothesis is
that by incorporating biophysical information, we can generate more
informative priors (models) and eventually reconstruct the ventricular
wall motion and the myocardial stresses more accurately.
We will discuss the formulation of the problem, the algorithmic and
parallel scalability of the numerical algorithms, and present
validation results on image datasets from subjects with normal and
abnormal cardiac function.
This is a joint work with Christos Davatzikos (University of
Pennsylvania), Harold Litt (University of Pennsylvania), Rahul Sampath
(Georgia Tech), and Hari Sundar (Siemens).
This page was made with 100% valid HTML & CSS - Send comments to Webmaster
Today's date and time is 12/19/12 - 16:48 CEST and this file (/conferences/aip2009/invited/talk.php) was last modified on 05/06/09 - 15:09 CEST