Gridbus Workflow Enactment Engine - Grid Computing: Infrastructure, Service, and Applications - page 134

diseased populations. These describe how the brain varies with age, gen-
der, and demographics. They can be used for identifying systematic effects
on brain structure. For instance, they provide a comprehensive approach
for studying a particular subgroup, with a specii c disease, receiving dif-
ferent medications, or neuropsychiatric disorder. Population-based atlases
contain anatomical models from many subjects. They store population
templates and statistical maps to summarize features of the population.
They also average individual images together so that common features of
the subgroup are reinforced.
Figure 5.16 shows a workl ow that employs the automated image regis-
tration (AIR) [18] and FSL [19] suite for creating population-based brain
atlases from high-resolution anatomical data. The stages of this workl ow
are as follows:

1. The inputs to the workl ow are a set of brain images, which are 3D
brain scans of a population with varying resolutions and a refer-
ence brain image. For each brain image, align_warp adjusts the
position and shape of each image to match the reference brain.
The output of each process is a warp parameter set dei ning the
spatial transformation to be performed.

fMRI images of different subjects

warp parameter sets

align_warp reslice

resliced images

atlas X slice

convert slicerX

atlas Y slice

softmean

convert slicerY

atlas
image

atlas Z slice

slicerZ

convert

generate 2D atlas

Parallel
processing

Sequential
processing

FIGURE 5.16 Population-based atlas workl ow.