Neuroimaging Data Processing/Processing/Steps/Skull Stripping

Neuroimaging Data Processing/Processing/Steps
Realignment Skull Stripping Tissue Segmentation

Concept

edit

Removal of the skull and other non-brain tissue like dura and eyes from anatomical images, which could otherwise complicate e.g. coregistration and normalization steps. Skull stripping can be part of the tissue segmentation (e.g. in SPM) but is mostly done by specialized algorithms that delineate the brain boundary. See [1] for a comparison of some brain extraction algorithms (BSE, BET, SPM, and McStrip), which suggests that all algorithms perform well in general but results highly depend on the particular dataset.

Implementation

edit

Skull stripping in SPM can be achieved by segmenting the anatomical scan, and using a thresholded version of the sum of grey and white matter probability maps to mask out the bias corrected structural scan.

This masking can be done using SPM --> Util --> Image Calculator (ImCalc button) directly in the batch editor or by pressing the ImCalc button on the SPM GUI. The image calculator allows to perform algebraic manipulations on a set of structural images.

 
Skull stripping in SPM ImCalc

Input Images = segmentation generated probability maps of grey matter(i1) and white matter(i2) and the original anatomical scan (i3)
Expression = defining the threshold to mask out the skull based on the selected tissue probability maps e.g ·[i3.*((i1 +i2) > 0.2) ] --> if you set the threshold higher then your skull strip is more stringent For older versions of SPM (or OldSeg), you can use the following expression: (i3.*(i1+i2))

BET and BET2

AFNI

edit

3dSkullStrip[2] is a program in AFNI to extract the brain from surrounding tissue from MRI T1-weighted images. The simplest command would be:

3dSkullStrip INPUTFILE

The process by default includes spatial normalization, some intensity normalization and repositioning of the brain, but to a certain extent those can be switched off. The actual skull stripping is a modified version of the BET[3] algorithm, expanding a spherical surface iteratively until it envelopes the brain. The output can be a skull stripped (masked) brain, the mask itself or different surface formats.

3dSkullStrip is called by a set of other afni functions like align_epi_anat.py, @auto_tlrc and 3dSeg. Therefore in afni_proc.py there is no separate block for skull-stripping but options can be adjusted as parts of the respective blocks.

References

edit
  1. Kristi Boesen, Kelly Rehm, Kirt Schaper, Sarah Stoltzner, Roger Woods, Eileen Lüders, David Rottenberg, Quantitative comparison of four brain extraction algorithms, NeuroImage, Volume 22, Issue 3, July 2004, Pages 1255-1261, ISSN 1053-8119, http://dx.doi.org/10.1016/j.neuroimage.2004.03.010.
  2. http://afni.nimh.nih.gov/pub/dist/doc/program_help/3dSkullStrip.html
  3. Smith, Stephen M., Fast robust automated brain extraction, Human Brain Mapping 2002, Volume 17 Issue 3, page 143-155, http://dx.doi.org/10.1002/hbm.10062