| |
Methods
74
This chapter mainly explains how this problem is solved. Information is given about the crea-
tion of atlases, the deformation procedure, the energy definition, the energy minimization, a
quantitative quality measure of the results, and about visualization.
5.2 Atlas Fitting
In mrFindBorders, an initially aligned atlas T is deformed to match the reference image R (see
Figure 35). Deforming the measured data to match the atlas is also possible. However, since it
is the aim of mrFindBorders to identify visual areas in the measured phase image, it would
have to be deformed back after the boundaries have been identified. Deforming the atlas to
match the data does not require that.
The deformation is computed by hierarchical spline-based image registration. A deriva-
tion and motivation of this method may be found in [Szeliski and Coughlan, 1994 #44]
[Szeliski and Coughlan, 1994 #45] [Szeliski and Shum, 1995 #46] [Szeliski and Shum, 1996
#47]. The procedure is explained in the following.
5.2.1 Creating Initially Aligned Atlas Images
Since different visual stimuli can be shown to a subject, the expected pattern of activity also
differs. Therefore, for different stimuli, different atlases have to be created. Figure 21 shows
an atlas for a rotating wedge visual stimulus, an atlas for a visual field sign map, and an atlas
for an expanding ring experiment. As explained before, a visual field sign map is calculated
by using functional images from both rotating wedge and expanding (or contracting) ring ex-
periments.
Based on points or a polygon specified by the user (see section 4.5), and pre-defined op-
tions, various parameters can be calculated by using simple geometric equations. The parame-
ters that are given to the function that generates the atlases for a rotating wedge stimulus, a
ring stimulus, and a sign map are listed next:
|  |
|
| |
|
|
