ants.segmentation.joint_label_fusion

joint_label_fusion(target_image, target_image_mask, atlas_list, beta=4, rad=2, label_list=None, rho=0.01, usecor=False, r_search=3, nonnegative=False, no_zeroes=False, max_lab_plus_one=False, output_prefix=None, verbose=False)[source]

A multiple atlas voting scheme to customize labels for a new subject. This function will also perform intensity fusion. It almost directly calls the C++ in the ANTs executable so is much faster than other variants in ANTsR.

One may want to normalize image intensities for each input image before passing to this function. If no labels are passed, we do intensity fusion. Note on computation time: the underlying C++ is multithreaded. You can control the number of threads by setting the environment variable ITK_GLOBAL_DEFAULT_NUMBER_OF_THREADS e.g. to use all or some of your CPUs. This will improve performance substantially. For instance, on a macbook pro from 2015, 8 cores improves speed by about 4x.

ANTsR function: jointLabelFusion

Parameters:

  • target_image (ants.core.ANTsImage) – image to be approximated

  • target_image_mask (ants.core.ANTsImage) – mask with value 1

  • atlas_list (list of ANTsImage types) – list containing intensity images

  • beta (scalar) – weight sharpness, default to 2

  • rad (scalar) – neighborhood radius, default to 2

  • label_list (list of ANTsImage types (optional)) – list containing images with segmentation labels

  • rho (scalar) – ridge penalty increases robustness to outliers but also makes image converge to average

  • usecor (bool) – employ correlation as local similarity

  • r_search (scalar) – radius of search, default is 3

  • nonnegative (bool) – constrain weights to be non-negative

  • no_zeroes (bool) – this will constrain the solution only to voxels that are always non-zero in the label list

  • max_lab_plus_one (bool) – this will add max label plus one to the non-zero parts of each label where the target mask is greater than one. NOTE: this will have a side effect of adding to the original label images that are passed to the program. It also guarantees that every position in the labels have some label, rather than none. Ie it guarantees to explicitly parcellate the input data.

  • output_prefix (str) – file prefix for storing output probabilityimages to disk

  • verbose (bool) – whether to show status updates

Returns:

segmentationANTsImage

segmentation image

intensityANTsImage

intensity image

probabilityimageslist of ANTsImage types

probability map image for each label

segmentation_numberslist of numbers

segmentation label (number, int) for each probability map

Return type:

dictionary w/ following key/value pairs

Example

>>> import ants
>>> ref = ants.image_read( ants.get_ants_data('r16'))
>>> ref = ants.resample_image(ref, (50,50),1,0)
>>> ref = ants.iMath(ref,'Normalize')
>>> mi = ants.image_read( ants.get_ants_data('r27'))
>>> mi2 = ants.image_read( ants.get_ants_data('r30'))
>>> mi3 = ants.image_read( ants.get_ants_data('r62'))
>>> mi4 = ants.image_read( ants.get_ants_data('r64'))
>>> mi5 = ants.image_read( ants.get_ants_data('r85'))
>>> refmask = ants.get_mask(ref)
>>> refmask = ants.iMath(refmask,'ME',2) # just to speed things up
>>> ilist = [mi,mi2,mi3,mi4,mi5]
>>> seglist = [None]*len(ilist)
>>> for i in range(len(ilist)):
>>>     ilist[i] = ants.iMath(ilist[i],'Normalize')
>>>     mytx = ants.registration(fixed=ref , moving=ilist[i] ,
>>>         type_of_transform = ('Affine') )
>>>     mywarpedimage = ants.apply_transforms(fixed=ref,moving=ilist[i],
>>>             transformlist=mytx['fwdtransforms'])
>>>     ilist[i] = mywarpedimage
>>>     seg = ants.threshold_image(ilist[i],'Otsu', 3)
>>>     seglist[i] = ( seg ) + ants.threshold_image( seg, 1, 3 ).morphology( operation='dilate', radius=3 )
>>> r = 2
>>> pp = ants.joint_label_fusion(ref, refmask, ilist, r_search=2,
>>>                     label_list=seglist, rad=[r]*ref.dimension )
>>> pp = ants.joint_label_fusion(ref,refmask,ilist, r_search=2, rad=[r]*ref.dimension)