Choosing structural sequences for your project:
- A quick T1 for visualization purposes only
- Extract fMRI signal from specific tissue or region
- A T2 for a better measure of sulcal CSF and/or true intracranial vault
- Measure volumes of gray matter, white matter, and CSF
- Measures of regional volumes
- Measures of the neocortex
For this simple purpose, a quick T1 volume can be acquired. If you have sufficient time (at least 7 minutes), it is highly recommended to acquire a relatively standard T1 of good quality to use for defining regions of interest. A T1 with better signal-to-noise ratio (SNR) and gray/white contrast will make it possible to measure brain structure directly (eg., the thickness of the neocortex)
Constrain fMRI activation signals to gray matter regions only: A single standard T1 can be used for this purpose, using segmentation tools such as FSL's FAST to define gray matter and white matter compartments. Note that if your participant has any abnormalities within the white matter, these regions can segment as gray matter in the resulting image. This is quite common in the normal aging brain, as well as in disorders such as Alzheimer's disease and HIV. The addition of a T2 scan, using multi-channel segmentation, provides a true measure of the CSF compartment as well.
Regions of Interest to extract fMRI signal Taking the gray matter mask a step further, a primary interest may be to extract functional signal from a specific region of interest that is more carefully aligned across participants, to account for morphometric variation due to individual differences and to disease states (e.g. enlarged ventricles). This application may require less accuracy than estimating volume or thickness directly, for example, removal of non-brain and careful boundary estimates may not have a significant impact given the differences in resolution of acquired datasets. Using careful image registration and T1-based tissue segmentation (e.g. FSL's FAST) to define gray matter, existing atlases (e.g. Harvard-Oxford atlas) can be applied to the gray matter mask to extract specific regional information.
When studying brain volume, ICV is an important variable used to control for individual differences in head size that often are unrelated to the specific scientific question at hand. ICV reflects the premorbid brain volume (as large as the brain ever was), providing a context that suggests how large the hippocampus should be (or should have been). ICV can also be useful as a direct dependent variable in studies where normal development has been altered.
A rough estimate of ICV can be made with a T1 using some available tools (e.g. FreeSurfer, this is an estimate derived from the atlas scaling factor on the basis of the transformation of the full brain mask into atlas space [Buckner et al., 2004]. For a more accurate estimate of true ICV, one should include all CSF, requiring a T2 because the T1 does not provide sulcal CSF information. That is, when the brain atrophies, it becomes smaller than the original ICV and tissue is replaced with CSF. Multi-channel T1 and T2 segmentation can be done using FSL's FAST.
The best approach to a three compartment segmentation, to include CSF, is a multi-channel tissue segmentation including both a T1 and a T2. A single, standard T1 can be used for estimates of gray and white matter compartments, using segmentation tools such as FSL's FAST, although segmentation is improved when including a T2 in the same approach. Note that if your participants have white matter disease, abnormalities can segment as gray matter (see Figure). This is common with normal aging and some disorders, and would result in an overestimation of gray matter volume and an underestimation of white matter volume.
Regional brain volumes can be estimated with a number of semi-automated programs based on a T1. Commonly estimated volumes that require quality review although little human intervention include the hippocampus, amygdala, caudate nucleus, putamen, thalamus, and the nucleus accumbens.
FSL's FIRST is a geometry-based approach that uses Bayesian shape and appearance models and relies on both shape and intensity priors, learned from a training set.
Freesurfer's segmentation approach employs a statistical, probabilistic atlas to assign anatomical labels to each voxel withing the brain mask [Fischl et al., 2002]. The classification process has a Bayesian basis that considers: 1) image intensity statistics calculated per anatomical label, 2) probablilties that a given anatomical label occurs at a given location, and 3) considerations of the labels assigned to neighboring voxels.
Various tools exist to measure volume, thickness, and surface area of the neocortex using a T1 volume. It is important to note that all estimates of the neocortex using a T1 are directly influenced by the adequacy of removal of non-brain ("skull stripping") since the cortical surface often lies adjacent to tissue that can be confused with brain tissue on a T1, such as meninges. Removal of non-brain can be particularly difficult requiring manual intervention and quality review is important to ensure appropriate estimate of cortical thickness. A common, freely available tool based on the reconstruction of the cortical surface is FreeSurfer. In addition to surface based analyses, regional cortical parcellation can be applied using existing atlases to assess volume, thickness, and surface area by region as well.