Neuroimaging Data Processing/Processing/Steps/Tissue Segmentation

Concept

Classification of the brain tissue into different tissue classes using high-resolution T1 weighted anatomical images. Usually gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) are separated but more classes are possible. Theoretically, in the T1-weighted image different tissue types can be separated based on their distinctive intensities. However, segmentation turns out much more thorny in reality because of noise that leads to overlapping intensity spectra of the different tissue types, partial volume effects i.e. single voxels containing different tissue types and remaining field inhomogeneities. Tissue segmentation algorithms can therefore also take into account tissue probability maps (e.g. in SPM) and geometric structure of the gray-white interface (e.g. FreeSurfer). For a comprehensive (though not so current) review of tissue segmentation methods see

Implementation

SPM

Segmentation New Segmentation

Unified segmentation in SPM is a one-click method that combines spatial normalization, bias field correction and tissue segmentation together. The prior probability that any voxel contains gray or white matter can be determined using a probabilistic atlas of tissue types; this prior probability is then combined with the data from the image determine the tissue class. Using this approach, two voxels with identical intensities can be identified as different tissue types.

FSL

FSL’s Brain Extraction Tool BET (Smith et al., 2004)

AFNI

3dSeg is an AFNI program for tissue segmentation. It allows for adding a variety of global and voxelwise priors but is not recommended for quantitative segmentation, e.g. for VBM. The most simple command (segmenting into CSF, GM and WM by default) is

3dSeg -anat ANATOMICAL DATASET

In afni_proc.py segmentation can be invoked through the (default) mask block via (apparently automatically implies -mask_segment_erode to calculate eroded tissue classes)

-mask_segment_anat yes

If WM and CSF should be used for tissue based regressors, there is an options which automates this process calles @ANATICOR. It is strongly recommended to use this in afni_proc.py where (in addition to the -mask_segment_anat like above) you would add

-regress_anaticor