The following can be used as a start for some text that describes how NVM can be used to segment MRI brain scans.
Structural MRI scans are segmented using open source software, "NVM" (freely
available from Neuromorphometrics, Inc. at http://neuromorphometrics.org:8080/nvm/).
Images are loaded by running NVM and providing a description of their location,
byte size, column, row, and slice dimensions and resolutions, along with other
parameters. The scan is then displayed in three orthogonal views and a pallet
of tools is used to adjust the display and segment the desired neuroanatomical
regions of interest. Scans can be positionally normalized to help decrease
measurement variability across multiple subjects by designating landmarks and
saving the re-sliced volume as a new scan. Three dimensional cropping can be
applied to make efficient use of display space.
Outlines around regions are created using isointensity contours along with
manual drawing and erasing. Intensities used to define isointensity contours
can be chosen by clicking on a location in the image or by taking a histogram
over a specific region and clicking on the histogram. Contours can then by
dynamically adjusted using the mouse. Manual editing is used to clean up and
join multiple contours. Contours are then "extracted" as outlines.
These outlines are assigned labels and saved for each slice where the desired
regions appear. Outlines can be shown filled in colors and/or toggled on and
off to facilitate review of their proper label assignment and precise boundary
location.
After segmentation is completed in this way, a menu option causes NVM to write
out a comma separated value (.csv) spreadsheet file that contains voxel counts
from all saved outline files. When loaded into Excel, volumes are calculated
in this spreadsheet by 1) multiplying the number of voxels enclosed in the
outlines by their volume and 2) adding half of the volume of the voxels located
on the outline itself. To make segmentation easier, NVM displays each slice
image at twice the original in-plane size so the row and column voxel dimensions
in the spreadsheet are half of their original values.