fnndsc/pl-mri-preview repository overview

⁠pl-mri-preview

pl-mri-preview is a ChRIS⁠ plugin to produce PNG image previews of the center slices of MRIs (NIFTI, MINC, ...) and also compute total brain volume.

⁠Installation

pl-mri-preview is a ChRIS⁠ plugin, meaning it can run from either within ChRIS or the command-line.

⁠Local Usage

To get started with local command-line usage, use Apptainer⁠ (a.k.a. Singularity) to run pl-mri-preview as a container:

singularity exec docker://fnndsc/pl-mri-preview mri_preview [--background 0.0] input/ output/

To print its available options, run:

singularity exec docker://fnndsc/pl-mri-preview mri_preview --help

⁠Examples

mri_preview requires two positional arguments: a directory containing input MRI images, and a directory where to create output PNGs.

mkdir incoming/ outgoing/
mv brain_recon.nii.gz incoming/
singularity exec docker://fnndsc/pl-mri-preview mri_preview incoming/ outgoing/

⁠File Types

⁠--inputs

Every input file with a file name ending with a value given by --inputs is processed. Unmatched files are ignored. Supported formats are listed on NiBabel⁠'s website.

⁠--units-fallback

pl-mri-preview can get voxel size units from the headers of NIFTI or MINC files. Units can be specified manually for other file types using --units-fallback.

https://github.com/nipy/nibabel/pull/1098⁠

⁠--outputs

pl-mri-preview creates image files. Supported output formats are any which are supported by matplotlib⁠, including .png, .jpg, and .svg.

The special type .txt writes a plaintext file instead, e.g.

2236612 voxels
230437.66977741406 mm^3

⁠About Brain Volume

⁠Input File

The input file should be masked (or it should be a mask), meaning it should have a background intensity which is lower than the foreground intensity -- usually the value denoting background is just 0 or 1.

pl-mri-preview is part of a pipeline which runs after pl-irtk-reconstruction⁠.

⁠v.s. mri_label_volume

mri_label_volume -a is a FreeSurfer tool to do the same thing, though their non-zero intensity threshold starts at and includes 5, which depending on what kind of input file you're working with, is probably an underestimation.

• https://github.com/freesurfer/freesurfer/blob/b11d2d547216a66dedeffc7b277feaf80c265795/mri_label_volume/mri_label_volume.cpp#L106-L113⁠
• https://github.com/freesurfer/freesurfer/blob/8c012da48fc79c99d5b315b169799c78847defa5/include/mri.h#L1138⁠
• https://github.com/freesurfer/freesurfer/blob/e34ae4559d26a971ad42f5739d28e84d38659759/utils/mriset.cpp#L3062-L3082⁠

To reproduce the behavior of mri_label_volume -a with pl-mri-preview, specify the value mri_preview --background 4.999.

⁠From Surfaces

Another computational approach to measure total brain volume is to compute the volume inside a topologically closed polygonal mesh representation of the pial matter. This object obtained from a surface extraction algorithm, such as ep-sphere_mesh⁠.

The advantage of measuring volume from a surface instead of in voxel space is that the computation is not discrete.