cteRunCrossSectional.sh

(Note: This page is automatically generated. Please do not attempt to edit it!)

This script runs cortical thickness estimation for a single subject, using voxel based methods.
In a nutshell, we run segmentation as described in [1], using linear [2] and non-linear [3]
registration, then run voxel based cortical thickness estimation as described in [4].

Usage: cteRunCrossSectional.sh image.nii mask.nii name [options] 

Mandatory Arguments:
 
  image.nii                : The T1 image you want the thickness estimation for.
  
  mask.nii                 : Brain mask, in same space as image.nii.
  
  name                     : Basename for the output. Output images are:
              
                             <name>_label.nii
                             <name>_GMfc.nii
                             <name>_WMfc.nii
                             <name>_CSFfc.nii
                             <name>_dGM.nii
                             <name>_iCSF.nii                         
                             <name>_thickness.nii
                             <name>_smoothed_thickness.nii

Options:

  -spm <directory>         : ** This will be compulsory if you don't already have spm in your MATLABPATH. **
                             SPM directory, which if specified is added onto the front of the MATLABPATH variable

  -tmproot <directory>     : Define a root directory for temporary files. Default /tmp
                             The difference between this option and the next, is that this
                             one takes the root directory (eg. /tmp) and creates a sub directory
                             with a process ID in it. (eg. /tmp/seg.1234) whereas the next option
                             just uses exactly what you give it.
                            
  -tmpdir <directory>      : Set temporary workspace directory.

  -keeptmp                 : Keep temporary workspace directory (for debugging purposes).
  
  -gpu                     : Turn on GPU for reg_aladin, reg_f3d. No memory checking done.

****************************
* Registration parameters  *
****************************

  -levels <int>            : The number of Free-Form Deformation multi-resolution levels in pyramid. Default 3.
  
  -levels_to_perform <int> : The number of Free-Form Deformation multi-resolution levels to actually optimise. Default 2.

  -dilations <int>         : Number of dilations to apply to mask, when registering/segmenting. Default 2.

****************************
* Segmentation parameters  *
****************************
  
  -atlas         image.nii : Atlas image, used for segmentation priors.
  
                             The atlas could, for example, be the
                             one found in spm8/canonical/avg152T1.nii or avg305T1.nii
                           
  -grey          image.nii : Grey matter prior, in same space as segmentation atlas. 
  -white         image.nii : White matter prior, in same space as segmentation atlas.
  -csf           image.nii : CSF prior, in same space as segmentation atlas.

                             These apriori arguments could, for example, be 
                             the priors in spm8/apriori directory.

  -deep_grey     image.nii : Grey matter prior image containing only deep grey matter,
                             in same space as segmentation atlas.
  
  -internal_csf  image.nii : CSF prior image, containing only internal CSF,
                             in same space as segmentation atlas.

****************************
* Labelling parameters     *
****************************

  -isotropic <float>       : Resample to isotropic voxels after segmentation step. Default off.
  
  -minp <float>            : When we classify GM, WM, CSF probability maps into a 3 label segmentation
                             you can specify a minimum probability (for GM only) below which the voxel is ignored.
                             This defaults to zero, and the classification picks the highest value
                             of the GM, WM, CSF as the label. 

  -connected               : Do connected component analysis when classifying GM, WM, CSF probability maps
                             into a 3 label segmentation. Default off.

  -refine_labels           : Refines the label image, by calling cteRefineLabelImage.sh. Default off.

****************************
* Thickness parameters     *
****************************
  
  -lapl_epsilon <float>    : Convergence tolerance for Laplacian solution.
  
  -lapl_iters <int>        : Maximum iterations for Laplacian solution.
  
  -pde_epsilon <float>     : Convergence tolerance for PDE solution (see [5]).
  
  -pde_iters <int>         : Maximum iterations for PDE solution (see [5]).
  
  -initp                   : Target probability threshold to iterate towards when doing Lagrangian initialisation. Default 0.5.

  -method [int]            : Slightly different methods in the cortical thickness bit:
                             1. Correct the GM, as describe in Acosta's paper, section 2.3, and use Lagrangian initialization
                             2. Correct the GM, as described in Bourgeat's paper, section 2.3.1, and use Lagrangian initialization
                             3. Do not correct the GM, and initialize the CSF/WM boundary voxels to minus half the voxel diagonal (images should really be isotropic), as described in Diep's paper
                             4. Use Jorge's method, no Lagrangian initialization.
                             5. Correct the GM, as described in Acosta's paper, then do Jorge's initialization, rather than Lagrangian

  -vmf [int]               : Voxel multiplication factor, a bit like supersampling. This overrides the -method flag.
                             So, if you are doing high res, the underlying method will be like specifying -method 1.
                                     
****************************
* Statistics parameters    *
****************************
  
  -region_atlas image.nii  : An image containing containing segmentation labels, in same space as atlas.

  -regions 1,2,3,4         : A comma separated list of region numbers to extract stats for.
                             These region numbers should correspond to the labels in your -stats_regions image.  
                             
  
[1] Manuel Jorge Cardoso, Matthew J. Clarkson, Gerard R. Ridgway,
Marc Modat, Nick C Fox and Sebastien Ourselin, 
"Improved Maximum A Posteriori Cortical Segmentation by Iterative Relaxation Of Priors" 
G.-Z. Yang et al. (Eds):MICCAI 2009, Part II, LNCS 5762, pp. 441-449, 2009.
                             
[2] Sebastien Ourselin, A Roche, G Subsol, Xavier Pennec, and Nicholas Ayache.
"Reconstructing a 3d structure from serial histological sections" 
Image and Vision Computing, 19(2000) 25-31: doi:10.1016/S0262-8856(00)00052-4    

[3] Marc Modat, Gerard R. Ridgway, Zeike A Taylor, Manja Lehmann,
Josephine Barnes, Nick C Fox, David J Hawkes, and Sebastien Ourselin.
"Fast free-form deformation using graphics processing units" 
Comput Meth Prog Bio 2009: doi:10.1016/j.cmpb.2009.09.002

[4] Oscar Acosta, Pierrick Bourgeat, Maria A. Zuluaga, Jurgen Fripp, Olivier Salvado
Sebastien Ourselin, and the Alzheimer's Disease NeuroImaging Initiative.
"Automated voxel-based 3D cortical thickness measurement in a combined Lagrangian-
Eulerian PDE approach using partial volume maps"
Medical Image Analysis 13 (2009) 730-743: doi:10.1016/j.media.2009.07.003

[5] Anthony. J. Yezzi and Jerry L. Prince.
"An Eulerian PDE approach for Computing Tissue Thickness"
IEEE Transactions On Medical Imaging Vol 22. No 10. October 2003.