Authors: Alexandre Coimbra, Jeffrey Saltzman and John Szumiloski
Affiliation: Merck Research Laboratories
Abstract
Magnetic Resonance Imaging has become a useful tool for a variety of tasks
in the pharmaceutical industry. These tasks range from understanding
clinical trial outcomes to the discovery of so-called biomarkers of disease
progression and/or drug efficacy. Often it is important to find a map or
coordinate transformation of an image to a reference image. This process is
called registration and can be set to yield a nonlinear mapping.
For example, in longitudinal studies the map is useful in finding changes in
morphology of important regions of the brain due to atrophy in Alzheimer?s
disease progression. In this presentation we describe a registration method
based on variational grid generation techniques. A variational integral is
proposed and the resulting Euler-Lagrange equations are three elliptic
partial differential equations. The equations are discretized and solved
using a full approximation scheme multigrid method. Both numerical results
and applications will be discussed.