| |
Flat Maps
21
nal fluid (CSF) are segmented first. Afterwards, the user selects the white matter of interest
and the white matter is verified by checking for handles and cavities.
Because gray matter surrounds white matter as a single sheet, gray matter segmentation
can be obtained from the white matter segmentation by growing gray matter from the white
matter boundary under certain constraints (Figure 7). This creates a connected representation
(mesh) of the gray matter, i.e., neighborhood relationships are known at the end. Contiguity of
the gray matter is determined by its relationship to the white matter. This ensures that two re-
gions of gray matter that are on different sides of a sulcus are not connected, even if they are
adjacent in the image, i.e., if they touch in the image. The contiguity of the gray matter is nec-
essary for the creation of accurate flattened representations of the gray matter [Wandell, 1999
#16]. This is the reason why gray matter is not segmented directly.
The described segmentation algorithm creates two result files. The classification file contains
the results of the automatic classification procedure as well as the hand edits (voxels in the
MR volume are labeled as white matter, CSF, and unknown). A second file contains informa-
Figure 7: Segmentation of gray matter from MRI images
A: The software automatically segments the white matter (tan overlay); B: The loca-
tion of gray matter is estimated by growing gray matter layers (green overlay) on the
white matter. This procedure ensures contiguity; Source: [Wandell, 1999 #16]
|  |
|
| |
|
|
