The Stereology and 3D Volume Analyses in Nervous Tissue - Animal Models of Acute Neurological Injuries: Injury and Mechanistic Assessments - page 46

patterns of gray matter loss in Alzheimer's disease were characterized
on the cortical surface ( 14 ). In this study, FreeSurfer (Table 3 ) was
used to extract and register cortical surfaces, and SurfStat (Table 3 )
was used to perform statistical analysis on the surface.
The SBM processing pipeline typically includes three steps:
(1) image segmentation to get the surfaces, (2) surface modeling
and alignment to model all the individual surfaces and register
them to an atlas so that surface data are comparable across objects,
and (3) extracting surface measurements (e.g., thickness, curvature,
etc.) and performing statistical inference on the surface. For
the fi rst step, Amira, FreeSurfer, or other segmentation tools can
be used to complete image segmentation, as described earlier.
For the second step, a surface modeling and alignment tool,
SPHARM-MAT (Table 3 , Fig. 10a ) ( 19, 20 ) will be used as an
example in order to demonstrate how this step is completed: (a)
Each input object is described as a 3D binary image (Fig. 10b ), which
is the result of image segmentation; (b) click the “Parameterization”
button to map the object surface onto the unit sphere so that
the object surface can be described by three spherical functions;
(c) click the “Expansion” button to expand each of these three
spherical functions into a linear combination of spherical harmonic
basis functions (i.e., the Fourier basis on the sphere) so that the
Fourier coeffi cients can be used to reconstruct the original sur-
face (Figs. 10c , d); (d) click the “Alignment” button to register
all the individual models to a template surface. Now all the surface
models are normalized and comparable across objects, and then
group analysis can be performed (see the third step). SPHARM-
MAT has other options for performing the above tasks; see its
documentation for details: http://imaging.indyrad.iupui.edu/projects/
SPHARM/ .
For the third step, as an example, SurfStat (Table 3 ) can be
used to perform statistical analysis of univariate and multivariate
surface data using linear mixed effects models and random fi eld
theory. Since both SPHARM-MAT and SurfStat are Matlab based;
it is straightforward to integrate both tools in SBM studies. For
example, one can fi rst use SPHARM-MAT to model and align sur-
faces and extract surface signals, and then use SurfStat for statistical
inference on the surface ( 21 ).
8. Summary
In the previous section, three popular processing pipelines have
been discussed for 3D analysis of neuroanatomical images: volu-
metric analysis, VBM, and SBM, where segmentation, registration,
and visualization are the major components implemented in these
pipelines. Typical protocols have been reviewed together with relevant