02. DARTEL Template Creation

This is an internal component, see Dartel pipeline for general information.

Paramètres

study_type: Choice ( input )

template_dir: DARTEL Template Dir ( sortie )

subjects_hierarchy: Save Subjects Hierarchy DISCO ( entrée )

Grey_DISCO: ListOf( Grey Mask Corrected DISCO ) ( input )

White_DISCO: ListOf( White Mask Corrected DISCO ) ( input )

dartel_space_TO_Talairach: Transform DARTEL Input Space to Talairach-AC/PC-Anatomist ( optional, entrée )

batch_location: Batch DARTEL Create Template ( sortie )

subjects_hierarchy_DARTEL: Save Subjects Hierarchy DARTEL ( sortie )

flow_field: ListOf( Flow Field DISCO DARTEL ) ( output )

regularisation_form: Choice ( input )

The registration is penalised by some "energy" term. Here, the form of this energy term is specified. Three different forms of regularisation can currently be used.

outer_iterations: Entier ( input )

Number of outer iterations (i.e. template generations) that DARTEL will generate. Each outer iteration consists of registering all subjects to the template from the previous outer iteration, then averaging the results to create the new template.

inner_iteration_1: Choice ( input )

The number of Gauss-Newton iterations to be done within this outer iteration. After this, new average(s) are created, which the individual images are warped to match.

regularisation_parameters_1: ListOf( Réel ) ( input )

For linear elasticity, the parameters are mu, lambda and id. For membrane energy, the parameters are lambda, unused and id.id is a term for penalising absolute displacements, and should therefore be small. For bending energy, the parameters are lambda, id1 and id2, and the regularisation is by (-lambda*Laplacian + id1)^2 + id2.
Use more regularisation for the early iterations so that the deformations are smooth, and then use less for the later ones so that the details can be better matched.

time_step_1: Choice ( input )

The number of time points used for solving the partial differential equations. A single time point would be equivalent to a small deformation model. Smaller values allow faster computations, but are less accurate in terms of inverse consistency and may result in the one-to-one mapping breaking down. Earlier iteration could use fewer time points, but later ones should use about 64 (or fewer if the deformations are very smooth).

smoothing_parameter_1: Choice ( input )

A LogOdds parameterisation of the template is smoothed using a multi-grid scheme. The amount of smoothing is determined by this parameter.

inner_iteration_2: Choice ( input )

regularisation_parameters_2: ListOf( Réel ) ( input )

time_step_2: Choice ( input )

smoothing_parameter_2: Choice ( input )

inner_iteration_3: Choice ( input )

regularisation_parameters_3: ListOf( Réel ) ( input )

time_step_3: Choice ( input )

smoothing_parameter_3: Choice ( input )

inner_iteration_4: Choice ( input )

regularisation_parameters_4: ListOf( Réel ) ( input )

time_step_4: Choice ( input )

smoothing_parameter_4: Choice ( input )

inner_iteration_5: Choice ( input )

regularisation_parameters_5: ListOf( Réel ) ( input )

time_step_5: Choice ( input )

smoothing_parameter_5: Choice ( input )

inner_iteration_6: Choice ( input )

regularisation_parameters_6: ListOf( Réel ) ( input )

time_step_6: Choice ( input )

smoothing_parameter_6: Choice ( input )

LM_Regularisation: Réel ( input )

Levenberg-Marquardt regularisation. Larger values increase the the stability of the optimisation, but slow it down. A value of zero results in a Gauss-Newton strategy, but this is not recommended as it may result in instabilities in the FMG.

cycles: Choice ( input )

Number of cycles used by the full multi-grid matrix solver. More cycles result in higher accuracy, but slow down the algorithm. See Numerical Recipes for more information on multi-grid methods.

iterations: Choice ( input )

Number of relaxation iterations performed in each multi-grid cycle. More iterations are needed if using "bending energy" regularisation, because the relaxation scheme only runs very slowly. See the chapter on solving partial differential equations in Numerical Recipes for more information about relaxation methods.

Informations techniques

Toolbox : Disco

Niveau d'utilisateur : 1

Identifiant : DartelTemplate

Nom de fichier : brainvisa/toolboxes/disco/processes/DARTEL_components/DartelTemplate.py

Supported file formats :

template_dir :

Répertoire, Répertoire

subjects_hierarchy :

Text file, Text file

Grey_DISCO :

NIFTI-1 image, NIFTI-1 image

White_DISCO :

NIFTI-1 image, NIFTI-1 image

dartel_space_TO_Talairach :

Transformation matrix, Transformation matrix

batch_location :

Matlab script, Matlab script

subjects_hierarchy_DARTEL :

Text file, Text file

flow_field :

NIFTI-1 image, NIFTI-1 image