rainygoblin / CTseg

Combined segmentation and spatial normalisation of computed tomography (CT) brain images

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

CTseg

This is a tool for segmenting and spatially normalising routine clinical, computed tomography (CT) brain scans. It is simillar to the widely used SPM software (see https://www.fil.ion.ucl.ac.uk/spm/software/spm12/), but with a number of key differences, for example:

  • The probabalistic atlas, which is deformed to the subject scan, has been learnt from a combination of MRIs and CTs and contain eight classes, representative of the tissue distribution of routine CT scans (see Fig. TODO):
    1. Background
    2. Soft tissue
    3. CSF
    4. WM
    5. GM
    6. Sagittal sinus and other
    7. Abnormality (e.g. a stroke lesion)
    8. Bone (skull and calcifications)
  • The deformation model combines affine with diffemorphic registration.
  • The gaussian mixture model now has priors on its means and covariances. The prior hyper-parameters has been learnt from a large number of CT scans, capturing their Hounsfield-based intensity distribution, for a more robust segmentation (see Fig. TODO).

The software takes as input either a NIfTI representation of the CT image (as a path), or the DICOMs (as a folder).

Usage

This tool runs in MATLAB and builds on the SPM12 software. SPM12 will therefore have to be downloaded (from https://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and put on the MATLAB path. Furthermore, make sure that both the Shoot and the Longitudinal toolbox of SPM12 (available in spm/toolbox/{Shoot,Longitudinal}) is also on the MATLAB path.

Having installed SPM12, add CTseg to SPM12 by copying it into the toolbox folder (spm/toolbox/CTseg). CTseg will now be available as a batch job in SPM12:

  1. Launch SPM12 from the MATLAB command window: spm pet
  2. Press Batch
  3. Press SPM -> Tools -> CT Segmentation
  4. Double click the CT scans option and pick one or more CT images.
  5. Press the green play button to run with default options (should work OK).

Depending what output you chose to write to disk, you will have this written to the Output directory option. By default this is ./CTseg-Results/image-name.

The possible outputs are:

  • Native space image
  • Template space image
  • Native space segmentations
  • DARTEL import segmentations
  • Template space segmentations
  • Modulated template space segmentations

and are set with the Write option.

Publication

A publication is currently being written up and will be added here once it has been published, for potential users of the software to cite.

About

Combined segmentation and spatial normalisation of computed tomography (CT) brain images

License:GNU General Public License v3.0


Languages

Language:MATLAB 92.5%Language:C++ 4.3%Language:C 3.2%