The aim of this study is the analysis of 3D morphological changes of human brains
after cerebral infarction. The change of brain volume caused by the infarction is
measured by high-dimensional transformations. These are determined by a multiresolution
full multigrid (FMG) movement model [4], resulting in a precise correlation of homologous
structures. Thus, volume changes are detected with subvoxel accuracy to determine
volume changes between the infarct and remote brain regions without manual interaction.
High-dimensional transformations are determined by application of the theory according
to Navier-Lamee. Volumes are modelled as an elastic medium. For complex deformations
the elastic model is extended to a movement model and the total spatial differences
are divided into discrete steps. The gray-value forces driving the movement of each
voxel in the source object are determined by minimizing the gray-value difference
between source and reference volume. The movement of each voxel is controlled by the
data itself and smoothed with respect to noise by the elasticity properties defined
by the Lamee parameters for each step. The numerical effort is overcome by a combined
multiresolution, full-multigrid method and the correlation of homologous structures
improved by application of the Scale-Space theory. To avoid rotational transformations,
coarse and fine alignment is achieved by the extended principle axes theory [2, 3]
and by a cross-correlation based procedure using a matrix-norm [1]. The MRI data sets
are acquired at different times, enabling the determination of morphological changes
over time. The application of the FMG model enables the investigation of morphological
brain changes over time. Furthermore, it is possible to detect ischemia-induced spatial
distortions and plastic alterations, as the changes of the volume are exactly determined
by the deformation fields. The results show, that even small infarcts result in morphological
changes not only adjacent to the ischemic area but also in remote areas and in some
cases the contralateral hemisphere. Supported by the Deutsche Forschungsgemeinschaft
(SFB 194 A6, A13, B2). References: [1] Schormann T, Dabringhaus A, Zilles K. Bioimaging 1993; 1: 19–28. [2] Schormann
T, Zilles K. IEEE TMI 1997; 16: 942–947. [3] Schormann T, Zilles K. Human Brain Mapping
1998a; 6: 339–347hf. [4] Schormann T. Method for computing and displaying 2D- and
3D-spatial differences of structures. Int. Patent PCT/EP99/04442 (1998b).
