labs.spatial_models.hroi

Module: labs.spatial_models.hroi

Inheritance diagram for nipy.labs.spatial_models.hroi:

This module contains the specification of ‘hierarchical ROI’ object, Which is used in spatial models of the library such as structural analysis

The connection with other classes is not completely satisfactory at the moment: there should be some intermediate classes between ‘Fields’ and ‘hroi’

AuthorBertrand Thirion, 2009-2011

Virgile Fritsch <virgile.fritsch@inria.fr>

Class

HierarchicalROI

class nipy.labs.spatial_models.hroi.HierarchicalROI(domain, label, parents, id=None)

Bases: SubDomains

Class that handles hierarchical ROIs

Parameters:

kint

Number of ROI in the SubDomains object

labelarray of shape (domain.size), dtype=np.int_

An array use to define which voxel belongs to which ROI. The label values greater than -1 correspond to subregions labelling. The labels are recomputed so as to be consecutive integers. The labels should not be accessed outside this class. One has to use the API mapping methods instead.

featuresdict {str: list of object, length=self.k}

Describe the voxels features, grouped by ROI

roi_featuresdict {str: array-like, shape=(self.k, roi_feature_dim)

Describe the ROI features. A special feature, id, is read-only and is used to give an unique identifier for region, which is persistent through the MROI objects manipulations. On should access the different ROI’s features using ids.

parentsnp.ndarray, shape(self.k)

self.parents[i] is the index of the parent of the i-th ROI.

TODO: have the parents as a list of id rather than a list of indices.

__init__(domain, label, parents, id=None)

Building the HierarchicalROI

copy()

Returns a copy of self.

self.domain is not copied.

feature_to_voxel_map(fid, roi=False, method='mean')

Convert a feature to a flat voxel-mapping array.

Parameters:

fid: str

Identifier of the feature to be mapped.

roi: bool, optional

If True, compute the map from a ROI feature.

method: str, optional

Representative feature computation method if fid is a feature and roi is True.

Returns:

res: array-like, shape=(domain.size, feature_dim)

A flat array, giving the correspondence between voxels and the feature.

get_coord(id=None)

Get coordinates of ROI’s voxels

Parameters:

id: any hashable type

Id of the ROI from which we want the voxels’ coordinates. Can be None (default) if we want all ROIs’s voxels coordinates.

Returns:

coords: array-like, shape=(roi_size, domain_dimension)

if an id is provided,

or list of arrays of shape(roi_size, domain_dimension)

if no id provided (default)

get_feature(fid, id=None)

Return a voxel-wise feature, grouped by ROI.

Parameters:

fid: str,

Feature to be returned

id: any hashable type

Id of the ROI from which we want to get the feature. Can be None (default) if we want all ROIs’s features.

Returns:

feature: array-like, shape=(roi_size, feature_dim)

if an id is provided,

or list of arrays, shape=(roi_size, feature_dim)

if no id provided (default)

get_id()

Return ROI’s id list.

Users must access ROIs with the use of the identifiers of this list and the methods that give access to their properties/features.

get_leaves_id()

Return the ids of the leaves.

get_local_volume(id=None)

Get volume of ROI’s voxels

Parameters:

id: any hashable type

Id of the ROI from which we want the voxels’ volumes. Can be None (default) if we want all ROIs’s voxels volumes.

Returns:

loc_volume: array-like, shape=(roi_size, ),

if an id is provided,

or list of arrays of shape(roi_size, )

if no id provided (default)

get_parents()

Return the parent of each node in the hierarchy

The parents are represented by their position in the nodes flat list.

TODO: The purpose of this class API is not to rely on this order, so we should have self.parents as a list of ids instead of a list of positions

get_roi_feature(fid, id=None)

get_size(id=None, ignore_children=True)

Get ROI size (counted in terms of voxels)

Parameters:

id: any hashable type, optional

Id of the ROI from which we want to get the size. Can be None (default) if we want all ROIs’s sizes.

ignore_children: bool, optional

Specify if the size of the node should include (ignore_children = False) or not the one of its children (ignore_children = True).

Returns:

size: int

if an id is provided,

or list of int

if no id provided (default)

get_volume(id=None, ignore_children=True)

Get ROI volume

Parameters:

id: any hashable type, optional

Id of the ROI from which we want to get the volume. Can be None (default) if we want all ROIs’s volumes.

ignore_childrenbool, optional

Specify if the volume of the node should include (ignore_children = False) or not the one of its children (ignore_children = True).

Returns:

volumefloat

if an id is provided,

or list of float

if no id provided (default)

integrate(fid=None, id=None)

Integrate certain feature on each ROI and return the k results

Parameters:

fidstr

Feature identifier. By default, the 1 function is integrated, yielding ROI volumes.

id: any hashable type

The ROI on which we want to integrate. Can be None if we want the results for every region.

Returns:

lsum = array of shape (self.k, self.feature[fid].shape[1]),

The results

make_forest()

Output an nipy forest structure to represent the ROI hierarchy.

make_graph()

Output an nipy graph structure to represent the ROI hierarchy.

merge_ascending(id_list, pull_features=None)

Remove the non-valid ROIs by including them in their parents when it exists.

Parameters:

id_list: list of id (any hashable type)

The id of the ROI to be merged into their parents. Nodes that are their own parent are unmodified.

pull_features: list of str

List of the ROI features that will be pooled from the children when they are merged into their parents. Otherwise, the receiving parent would keep its own ROI feature.

merge_descending(pull_features=None)

Remove the items with only one son by including them in their son

Parameters:

methods indicates the way possible features are dealt with

(not implemented yet)

Notes

Caveat: if roi_features have been defined, they will be removed

plot_feature(fid, ax=None)

Boxplot the distribution of features within ROIs. Note that this assumes 1-d features.

Parameters:

fid: string

the feature identifier

ax: axis handle, optional

recompute_labels()

Redefine labels so that they are consecutive integers.

Labels are used as a map to associate voxels to a given ROI. It is an inner object that should not be accessed outside this class. The number of nodes is updated appropriately.

Notes

This method must be called every time the MROI structure is modified.

reduce_to_leaves()

Create a new set of rois which are only the leaves of self.

Modification of the structure is done in place. One way therefore want to work on a copy a of a given HROI object.

remove_feature(fid)

Remove a certain feature

Parameters:

fid: str

Feature id

Returns:

fobject

The removed feature.

remove_roi_feature(fid)

Remove a certain ROI feature.

The id ROI feature cannot be removed.

Returns:

fobject

The removed Roi feature.

representative_feature(fid, method='mean', id=None, ignore_children=True, assess_quality=True)

Compute a ROI representative of a given feature.

Parameters:

fid: str,

Feature id

method: str,

Method used to compute a representative. Chosen among ‘mean’ (default), ‘max’, ‘median’, ‘min’, ‘weighted mean’.

id: any hashable type

Id of the ROI from which we want to extract a representative feature. Can be None (default) if we want to get all ROIs’s representatives.

ignore_children: bool,

Specify if the volume of the node should include (ignore_children = False) or not the one of its children (ignore_children = True).

assess_quality: bool

If True, a new roi feature is created, which represent the quality of the feature representative (the number of non-nan value for the feature over the ROI size). Default is False.

select_id(id, roi=True)

Convert a ROI id into an index to be used to index features safely.

Parameters:

idany hashable type, must be in self.get_id()

The id of the region one wants to access.

roibool

If True (default), return the ROI index in the ROI list. If False, return the indices of the voxels of the ROI with the given id. That way, internal access to self.label can be made.

Returns:

indexint or np.array of shape (roi.size, )

Either the position of the ROI in the ROI list (if roi == True), or the positions of the voxels of the ROI with id id with respect to the self.label array.

select_roi(id_list)

Returns an instance of HROI with only the subset of chosen ROIs.

The hierarchy is set accordingly.

Parameters:

id_list: list of id (any hashable type)

The id of the ROI to be kept in the structure.

set_feature(fid, data, id=None, override=False)

Append or modify a feature

Parameters:

fidstr

feature identifier

data: list or array

The feature data. Can be a list of self.k arrays of shape(self.size[k], p) or array of shape(self.size[k])

id: any hashable type, optional

Id of the ROI from which we want to set the feature. Can be None (default) if we want to set all ROIs’s features.

override: bool, optional

Allow feature overriding

Note that we cannot create a feature having the same name than

a ROI feature.

set_roi_feature(fid, data, id=None, override=False)

Append or modify a ROI feature

Parameters:

fid: str,

feature identifier

data: list of self.k features or a single feature

The ROI feature data

id: any hashable type

Id of the ROI of which we want to set the ROI feature. Can be None (default) if we want to set all ROIs’s ROI features.

override: bool, optional,

Allow feature overriding

Note that we cannot create a ROI feature having the same name than

a feature.

Note that the `id` feature cannot be modified as an internal

component.

to_image(fid=None, roi=False, method='mean', descrip=None)

Generates a label image that represents self.

Parameters:

fid: str,

Feature to be represented. If None, a binary image of the MROI domain will be we created.

roi: bool,

Whether or not to write the desired feature as a ROI one. (i.e. a ROI feature corresponding to fid will be looked upon, and if not found, a representative feature will be computed from the fid feature).

method: str,

If a feature is written as a ROI feature, this keyword tweaks the way the representative feature is computed.

descrip: str,

Description of the image, to be written in its header.

Returns:

nimnibabel nifti image

Nifti image corresponding to the ROI feature to be written.

Notes

Requires that self.dom is an ddom.NDGridDomain

Functions

nipy.labs.spatial_models.hroi.HROI_as_discrete_domain_blobs(domain, data, threshold=-inf, smin=0, criterion='size')

Instantiate an HierarchicalROI as the blob decomposition of data in a certain domain.

Parameters:

domaindiscrete_domain.StructuredDomain instance,

Definition of the spatial context.

dataarray of shape (domain.size)

The corresponding data field.

thresholdfloat, optional

Thresholding level.

criterionstring, optional

To be chosen among ‘size’ or ‘volume’.

smin: float, optional

A threshold on the criterion.

Returns:

nroi: HierachicalROI instance with a signal feature.

nipy.labs.spatial_models.hroi.HROI_from_watershed(domain, data, threshold=-inf)

Instantiate an HierarchicalROI as the watershed of a certain dataset

Parameters:

domain: discrete_domain.StructuredDomain instance

Definition of the spatial context.

data: array of shape (domain.size)

The corresponding data field.

threshold: float, optional

Thresholding level.

Returns:

nroiHierarchichalROI instance

The HierachicalROI instance with a seed feature.

nipy.labs.spatial_models.hroi.hroi_agglomeration(input_hroi, criterion='size', smin=0)

Performs an agglomeration then a selection of regions so that a certain size or volume criterion is satisfied.

Parameters:

input_hroi: HierarchicalROI instance

The input hROI

criterion: str, optional

To be chosen among ‘size’ or ‘volume’

smin: float, optional

The applied criterion

Returns:

output_hroi: HierarchicalROI instance

nipy.labs.spatial_models.hroi.make_hroi_from_subdomain(sub_domain, parents)

Instantiate an HROi from a SubDomain instance and parents