VoxelBased Morphometry VBM

1
Voxel-Based Morphometry(VBM)

• Brian T. Cabaniss
• Cowan Lab
• July 9, 2007

2
VBM basics

• Utilizes structural MRI images
• Unbiased, whole brain technique
• The output of VBM can be either information
  concerning regional volume or tissue
  concentration (density)
• The output can be displayed as a statistical
  parametric map.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
3
VBM basics

• VBM is able to detect smaller changes in brain
  volume than is possible using a region of
  interest (ROI) approach.
• VBM has been designed to be sensitive to
  differences in local compositions of various
  brain tissue types, such as gray matter.
• VBM is not sensitive to differences among
  individuals that are due to volume and position
  down to a specified scale.

Ashburner and Friston, Why Voxel-Based
Morphometry Should Be Used (2001)
4
Preprocessing
Standard Protocol
Optimized Protocol
5
Preprocessing
Mechelli et. al., Voxel-Based Morphometry of the
Human Brain Methods and Applications (2005)
6
Preprocessing

• Function to shape the data in such a way that
  makes statistical analysis sensitive for local
  changes in tissue composition.
• 3 General Steps for Preprocessing a T1 Image for
  Optimized VBM
• Segment
• Spatially Normalize
• Smooth
• The optimized procedure also involves modulating
  the data to yield volume information.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
7
Segmentation

• Segmentation is an automated process that
  seperates tissue types with mixture model cluster
  analysis based on
• 1. Voxel intensities
• 2. A priori knowledge of the location of gray
  matter, white matter, CSF, and other tissues in
  normal brains
• and is responsible for labeling and extracting
  such things as -Gray matter -White
  matter -CSF -Other (skull, dural venous sinus,
  etc.)

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
8
Segmentation

• The first round of segmentation takes the T1
  images in native space and is used to create
• -Gray matter images in native space
• -White matter images in native space

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
9
Spatial Normalization

• Normalization corrects for global differences in
  position.
• Normalization is a transformation to stereotactic
  space.
• Each gray matter image is normalized to a gray
  matter template, for example, MNI or a template
  created by the investigator from study-specific
  subject data.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
Ashburner and Friston, Why Voxel-Based
Morphometry Should Be Used (2001)
Mechelli et. al., Voxel-Based Morphometry of the
Human Brain Methods and Applications (2005)
10
Spatial Normalization

• Normalization occurs as follows
• 1. Estimate the best 12-parameter affine
  transformation. Aided by a maximum a posteriori
  estimate of normal variability in brain size.
• 2. Correction for non-linear, global
  differences. This correction is constructed by a
  linear combination of smooth spatial basis
  functions.
• 3. A mask weights the normalization to brain
  instead of non-brain.
• 4. Reslice the images.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
11
Apply Normalization Parameters

• Use the normalization parameters gained from
  normalizing the gray matter or white matter to a
  template and apply them to the original T1
  images.
• Segmentation is reapplied. This second
  segmentation step is to clear possibly remaining
  nonbrain voxels.

Mechelli et. al., Voxel-Based Morphometry of the
Human Brain Methods and Applications (2005)
Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
12
Smoothing

• Modifying the data to fit a certain distribution
  is necessary for statistical parametric analysis
  to be valid.
• In the case of VBM, the data must be normally
  distributed as a Gaussian field model is used for
  statistical analysis.
• Smoothing is performed with a user defined
  smoothing kernel. An 8-mm FWHM isotropic
  Gaussian kernel is often employed.
• Smoothing causes the value of each voxel to be
  the average value of its neighboring voxels.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
13
Smoothing

• Smoothing with a FWHM isotropic Gaussian kernel
  inherently makes the data more normally
  distributed by the central limit theorem.
• Central Limit Theorem the summation of many
  variables which have a finite variance will
  produce a sum that is approximately normally
  distributed.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
14
Example of Normalizing Data
http//asymptote.sourceforge.net/doc/histogram.png
15
FWHM
16
Optimized vs Standard VBM

• Optimized VBM removes the missegmentation that is
  sometimes seen in Standard VBM through the second
  segmentation step.
• Optimized VBM also employs a modulation step.
• Nonlinear spatial normalization during
  preprocessing causes brain regions to
  differentially experience a change in volume.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
17
Optimized vs Standard VBM

• Modulation (voxel values) x (Jacobian
  determinants) (reestablishing volume
  information)
• Thus, the output of Optimized VBM will be
  information about percentage of brain volume.
  For example, Group As brain structure X is 5 of
  their total brain volume whereas Group Bs brain
  structure X is 7 of their total brain volume.
• You do not get information on absolute volume
  size, such as brain structure X is .5 ccs.
• The output of Standard VBM is tissue
  concentration, or in other words, the proportion
  of the type of tissue, such as gray matter, to
  the proportion of all other tissue types in the
  given region.

Good et. al., A Voxel-Based Morphometric Study of
Ageing in 465 Normal Adult Human Brains (2001)
Mechelli et. al., Voxel-Based Morphometry of the
Human Brain Methods and Applications (2005)
http//en.wikibooks.org/wiki/SPM-VBM
18
SPM

• The final step is to utilize SPM to create
  statistical parametric maps.