sdcflows.utils.tools module

Image processing tools.

sdcflows.utils.tools.brain_masker(in_file, out_file=None, padding=5)[source]

Use grayscale morphological operations to obtain a quick mask of EPI data.

sdcflows.utils.tools.deoblique_and_zooms(in_reference: SpatialImage, oblique: SpatialImage, factor: int = 4, padding: int = 1, factor_tol: float = 0.0001)[source]

Generate a sampling reference aligned with in_reference fully covering oblique.

Parameters:
  • in_reference (SpatialImage) – The sampling reference.

  • oblique (SpatialImage) – The rotated coordinate system whose extent should be fully covered by the generated reference.

  • factor (int) – A factor to increase the resolution of the generated reference.

  • padding (int) – Number of additional voxels around the most extreme positions of the projection of oblique on to the reference.

  • factor_tol (float) – Absolute tolerance to determine whether factor is one.

sdcflows.utils.tools.ensure_positive_cosines(img)[source]

Reorient axes polarity to have all positive direction cosines.

In other words, this interface will reorient the image polarities to be all positive, respecting the axes ordering. For instance, LAS -> RAS or ALS -> ARS.

sdcflows.utils.tools.reorient_pedir(pe_dir, source_ornt, target_ornt=None)[source]

Return updated PhaseEncodingDirection to account for an image data array rotation

Orientations form a natural group with identity, product and inverse. This function thus has the following properties (here e is the identity, * the product and - the inverse; a and b are arbitrary orientations):

reorient(pe_dir, e, e) == pe_dir reorient(pe_dir, a, e) == reorient(pe_dir, a) reorient(pe_dir, e, b) == reorient(pe_dir, -b) reorient(pe_dir, a, b) == reorient(pe_dir, a * -b, e)

Therefore, this function accepts one or two orientations, and precomputed transforms from a to b can be passed as the “source” orientation.

Passing only a source orientation will rotate into RAS:

>>> from nibabel.orientations import axcodes2ornt
>>> reorient_pedir("j", axcodes2ornt("RAS"))
'j'
>>> reorient_pedir("i", axcodes2ornt("PSL"))
'j-'

Passing only a target_ornt will rotate from RAS:

>>> reorient_pedir("j", source_ornt=None, target_ornt=axcodes2ornt("RAS"))
'j'
>>> reorient_pedir("i", source_ornt=None, target_ornt=axcodes2ornt("PSL"))
'k-'

Passing both will rotate from source to target.

>>> reorient_pedir("j", axcodes2ornt("LPS"), axcodes2ornt("AIR"))
'i-'

Passing a transform orientation as source_ornt will perform the transform, and passing it as target_ornt will invert the transform:

>>> from nibabel.orientations import ornt_transform
>>> xfm = ornt_transform(axcodes2ornt("LPS"), axcodes2ornt("AIR"))
>>> reorient_pedir("j", xfm)
'i-'
>>> reorient_pedir("j", source_ornt=None, target_ornt=xfm)
'k-'
sdcflows.utils.tools.resample_to_zooms(in_file, zooms, order=3, prefilter=True)[source]

Resample the input data to a new grid with the requested zooms.