Skull-Strip MRI

This process provides basic skull-stripping algorithms for T1- and T2-weighted MRIs. They are based on a first k-means segmentation and morphomathematical operations. They also include steps dedicated to preclinical imaging, in particular to deal with the massive head muscles that surround the brain of rodents and non-human primates.

Description

T1-weighted MRI

The T1 algorithm starts with a 2-class k-means (equivalent to Otsu's thresholding). An erosion is applied to the obtained mask in order to separate the brain from the surrounding tissues. The brain component is selected as the fillest one, where the filling coefficient is computed as the ratio between the component volume and its bounding box volume. The complete brain mask is then reconvered with a dilation of this component into a constrained space. Finally, a closing operation is applied.

Several parameters can modify the resulting mask. The most useful is exclude_biggest which, if activated, will exclude the biggest component when selecting the brain component. Usually, our filling ratio is able to separate the muscle component from the brain component. However, in some cases, it fails. In such cases, excluding the biggest component should help obtaining an accurate brain mask.

In expert mode, a few other parameters can be changed:

• k: the number of classes in the k-means segmentation.
• radius_erosion: the radius of the erosion used to separate components (mm).
• radius_dilation: the radius of the dilation used to recover tissues (mm).
• radius_closing: the radius of the final closing operation (mm).

T2-weighted MRI

In T2-weighted images, muscles produce an hyposignal compared to brain tissues. The skull-stripping process is thus simplified. A 6-class k-means segmentation is performed with the two more hypo-intense classes classified as background and muscle, the two more hyper-intense classes classified as CSF, and the two middle classes classifed as brain tissue.

Most background classes are first excluded through a series of morphomathematical operations applied to the brain + csf mask (Filling, Erosion, Biggest, Filling, Dilation). Spurious tissues are then removed through an other series of operations applied to the brain mask (Erosion, Biggest, Filling, Dilation). Holes are then filled through a final series of operations (Closing, Filling).

In advanced mode, some parameters can be modulated in order to deal with failing cases:

• nclasses: number of classes in the k-means segmentation.
• nbackground: number of most hypointense classes that should be classified as background.
• ncsf: number of most hyperintense classes that should be classified as CSF.
• radius: radius of the morphomathematical operations (mm).

Parameters

input_mri_nobias: P:MRI Bias Corrected ( input )

Input bias-corrected MRI. A previous bias correction is mandatory in order for the k-means segmentation to work accurately.

contrast: OpenChoice ( optional, input )

Contrast of the MRI (T1- or T2-weighted).

specie: OpenChoice ( optional, input )

Specie processed.

nclasses: Integer ( optional, input )

[T2] Number of classes in the k-means segmentation.

nbackground: Integer ( optional, input )

[T2] Number of most hypointense classes that should be classified as background.

ncsf: Integer ( optional, input )

[T2] Number of most hyperintense classes that should be classified as CSF.

radius: Float ( optional, input )

[T2] Radius of the morphomathematical operations (mm).

k: Integer ( optional, input )

[T1] Number of classes in the k-means segmentation.

radius_erosion: Float ( optional, input )

[T1] Radius of the erosion used to separate components (mm).

radius_dilation: Float ( optional, input )

[T1] Radius of the dilation used to recover tissues (mm).

radius_closing: Float ( optional, input )

[T1] Radius of the final closing operation (mm).

exclude_biggest: Boolean ( optional, input )

Exclude the biggest component when selecting the brain component. Usually, our filling ratio is able to separate the muscle component from the brain component. However, in some cases, it fails. In such cases, excluding the biggest component should help obtaining an accurate brain mask.

output_skullstrip_mask: P:Morpho Skull Stripping Mask ( optional, output )

Output brain mask that should contain brain tissues + CSF.

dbg_visu_steps: Boolean ( optional, input )

If activated, intermediate steps will be shown during the processing.

Technical information

Toolbox : Primatologist

User level : 0

Identifier : primate_MaskForRegistration

File name : brainvisa/toolboxes/primatologist/processes/blocks/primate_MaskForRegistration.py

Supported file formats :

input_mri_nobias :

gz compressed NIFTI-1 image, Aperio svs, BMP image, DICOM image, Directory, ECAT i image, ECAT v image, FDF image, FreesurferMGH, FreesurferMGZ, GIF image, GIS image, Hamamatsu ndpi, Hamamatsu vms, Hamamatsu vmu, JPEG image, Leica scn, MINC image, NIFTI-1 image, PBM image, PGM image, PNG image, PPM image, SPM image, Sakura svslide, TIFF image, TIFF image, TIFF(.tif) image, TIFF(.tif) image, VIDA image, Ventana bif, XBM image, XPM image, Zeiss czi, gz compressed MINC image, gz compressed NIFTI-1 image

output_skullstrip_mask :

gz compressed NIFTI-1 image, BMP image, DICOM image, Directory, ECAT i image, ECAT v image, FDF image, GIF image, GIS image, JPEG image, MINC image, NIFTI-1 image, PBM image, PGM image, PNG image, PPM image, SPM image, TIFF image, TIFF(.tif) image, VIDA image, XBM image, XPM image, gz compressed MINC image, gz compressed NIFTI-1 image