Brain Parcellation (EM)

Description

References

• Y. Balbastre, "Morphometry analysis tools for the Macaque brain: development and validation," PhD thesis, Université Paris-Saclay, 2016.

Parameters

analysis: String ( optional, input )

Name of the analysis attribute that should be attached to this pipeline output files. If not provided, decision is let to the auto filling system.

input_mri: P:MRI Raw ( input )

Raw MRI.

input_mask: P:Atlas Registered Skull Stripping Mask ( optional, input )

Skull-stripping mask. All subsequent analyses will be restricted to voxels inside this mask (parameter estimation, segmentation...). It will also be used to lower the memory load by only loading data inside the mask bounding box.

input_bias_field: P:MRI VIP Bias Field ( optional, input )

Bias field initialization. If provided, the bias field will be initialized with this image before EM refinement.

input_prior: P:EM Prior Proba ( optional, input )

Non-stationary prior probabilities. This 4D volume should contain a priori probabilities to observe a given class at a any location.

input_clique_matrix: P:Atlas Clique Matrix ( optional, input )

Matrix containing a priori probabilities on cliques, i.e. possible neighbouring configurations.

input_hierarchy: P:Atlas Labels Hierarchy ( optional, input )

Hierarchy associated with the atlas used.

input_initial_param: P:Complete Gaussian Mixture Parameters ( optional, input )

Initial values for the Gaussian mixture (mean, sd, proportion).

contrast: OpenChoice ( optional, input )

is_mapping: Boolean ( optional, input )

Set to true if the input image is a parametric map (T1 mapping, T2 mapping, etc) instead of a magnitude T1 or T2 weighted image. If false,input data will be log-transformed for the mixture fitting.

nb_classes: Integer ( optional, input )

Number of mixture classes.

initial_param: Choice ( optional, input )

Method used to initialize mixture parameters:

• Provided: values are provided, either with a json file (input_initial_param), or by manually entering them (input_mu, input_sigma, input_alpha).
• Random: values will be randomly set.
• From Prior: maximum-likelihod estimations using the provided non-stationary prior will be used.
• KMeans: a first classification by k-means will be used to initialize the parameters.
• KMeans++: a first classification by a robust version of the k-means algorithm will be used to initialize the parameters.

initial_mu: ListOf( Float ) ( optional, input )

Initial values for the mixture means.

initial_sigma: ListOf( Float ) ( optional, input )

Initial values for the mixture standard deviations.

initial_alpha: ListOf( Float ) ( optional, input )

Initial values for the mixture proportions.

parameterize_prior: Boolean ( optional, input )

If true, the proportion of voxels in each mixture class will be optimized by EM. If true and non-stationary priors were provided, it is the ratio between voxels proportions in the model and in the prior that will be optimized.

prior_type: Choice ( optional, input )

If stationary, prior probabilitiy of observing a class at a given location is the same in the whole image. If non-stationary, it is different in each location.

smooth_prior: OpenChoice ( optional, input )

Should we perform an additional smoothing of the non-stationary prior (it might be useful if the image aspect is very unusual, pathological, etc.).

stop_criterion: Choice ( optional, input )

Criterion used to stop the EM optimization.

stop_value: Float ( optional, input )

Criterion value under which the EM optimization is stopped.

iter_min: Integer ( optional, input )

Minimum number of EM iterations.

iter_max: OpenChoice ( optional, input )

Maximum number of EM iterations. If Auto, it is automatically computed as 100/nb_classes, so that processing time is lowered in the case of large mixtures.

denoising: Boolean ( optional, input )

Estimate noise during EM optimization?

bias_correction: Choice ( optional, input )

Bias correction type.

bias_grid: ListOf( Float ) ( optional, input )

Resolution of the bias grid (bias_unit should also be checked).

bias_unit: Choice ( optional, input )

Unit of the bias grid resolution.

exclude_from_bias: ListOf( String ) ( optional, input )

Classes (names or labels), that should be excluded when estimating the bias field. Because the MR signal in the background and in the CSF is often non-Gaussian, it is better to excluded them.

random_field: Choice ( optional, input )

When should we use MRF regularization:

• No: no MRF is used. The model is then that of a common Gaussian mixture, with independant voxels.
• At each EM iteration: common Gaussian MRF.
• At the end only: A common Gaussian mixture is used during EM optimization of the model parameters, but the MRF is used to compute the final posterior and segmentation. This can be useful to speed up computation.

rf_type: Choice ( optional, input )

• Markov: the MRF posterior is defined as P( x{i} | x{Ni} ) = Prod{j in Ni} C(x{i}, x{j})^beta, where C is the clique matrix.
• Gibbs: the MRF posterior is defined as P( x{i} | x{Ni} ) = exp( beta * Sum{j in Ni} C(x{i}, x{j}) ), where C is the clique matrix.

rf_order: Choice ( optional, input )

Connectivity order of the MRF lattice.

rf_eval_mode: Choice ( optional, input )

• Iterated Conditional Modes: the current best estimate of the segmentation is used to compute the MRF posterior until convergence.
• Maximum Marginal Proabilities: in the MRF posterior, neighbouring probabilities are approximated with that of a common mixture model (without MRF).

icm_iter_min: Integer ( optional, input )

Minimum number of iterations for ICM.

icm_iter_max: Integer ( optional, input )

Maximum number of iterations for ICM.

beta: OpenChoice ( optional, input )

Parameter used to modulate clique priors. If beta = 1, clique priors are kept as is. If beta = 0, it i equivalent to a common mixture model (withour MRF). The bigger the beta, the more compact the resulting regions. By default, the ratio between the atlas and the target image resolutions is used.

write_bias_field: Boolean ( optional, input )

Write the EM estimated bias field.

write_mri_nobias: Boolean ( optional, input )

Write the unbiased image.

write_parcellation: Boolean ( optional, input )

Write the resulting segmentation.

write_proba_parcellation: Boolean ( optional, input )

Write the posterior probabilities. They are more accurate than the segmentation to compute volumetry measures.

output_bias_field: P:MRI EM Bias Field ( output )

EM-optimized bias field.

output_mri_nobias: P:MRI EM Bias Corrected ( output )

Unbiased image.

output_parcellation: P:EM Parcellation ( output )

Resulting segmentation.

output_proba_parcellation: P:EM Parcellation Proba ( output )

Posterior probabilities.

verbose: Integer ( optional, input )

Make the process more verbose.

debug: Directory ( optional, output )

If the path to a folder is provided, intermediate volumes will be written there.

Technical information

Toolbox : Primatologist

User level : 0

Identifier : primate_Parcellation

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

Supported file formats :

input_mri :

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

input_mask :

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

input_bias_field :

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

input_prior :

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

input_clique_matrix :

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

input_hierarchy :

Hierarchy, Hierarchy

input_initial_param :

JSON file, JSON file

output_bias_field :

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

output_mri_nobias :

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

output_parcellation :

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

output_proba_parcellation :

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

debug :

Directory, Directory