wsireg.reg_transforms package

Submodules

wsireg.reg_transforms.reg_transform module

class wsireg.reg_transforms.reg_transform.RegTransform(elastix_transform)

Bases: object

Container for elastix transform that manages inversion and other metadata. Converts elastix transformation dict to it’s SimpleITK representation

elastix_transform

elastix transform stored in a python dict

Type:

dict

itk_transform

elastix transform in SimpleITK container

Type:

sitk.Transform

output_spacing

Spacing of the targeted image during registration

Type:

list of float

output_size

Size of the targeted image during registration

Type:

list of int

output_direction

Direction of the targeted image during registration (not relevant for 2D applications)

Type:

list of float

output_origin

Origin of the targeted image during registration

Type:

list of float

resampler_interpolator

elastix interpolator setting for resampling the image

Type:

str

is_linear

Whether the given transform is linear or non-linear (non-rigid)

Type:

bool

inverse_transform

Inverse of the itk transform used for transforming from moving to fixed space Only calculated for non-rigid transforms when called by compute_inverse_nonlinear as the process is quite memory and computationally intensive

Type:

sitk.Transform or None

as_np_matrix(use_np_ordering: bool = False, n_dim: int = 3, use_inverse: bool = False, to_px_idx: bool = False) → ndarray | None

Creates a affine transform matrix as np.ndarray whether the center of rotation is 0,0. Optionally in physical or pixel coordinates. :param use_np_ordering: Use numpy ordering of yx (napari-compatible) :type use_np_ordering: bool :param n_dim: Number of dimensions in the affine matrix, using 3 creates a 3x3 array :type n_dim: int :param use_inverse: return the inverse affine transformation :type use_inverse: bool :param to_px_idx: return the transformation matrix specified in pixels or physical (microns) :type to_px_idx: bool

Returns:

full_matrix – Affine transformation matrix

Return type:

np.ndarray

compute_inverse_nonlinear() → None

Compute the inverse of a BSpline transform using ITK

wsireg.reg_transforms.reg_transform_seq module

class wsireg.reg_transforms.reg_transform_seq.RegTransformSeq(reg_transforms: str | Path | Dict[str, List[str]] | None = None, transform_seq_idx: List[int] | None = None)

Bases: object

Class to concatenate and compose sequences of transformations

add_transforms(transforms: str | Path | dict | List[RegTransform] | RegTransform, transform_seq_idx: List[int] | None = None) → None

Add transforms to sequence.

Parameters:

• transforms (path to wsireg transforms .json, elastix transform dict,RegTransform ot List of RegTransform) –

• transform_seq_idx (list of int) – Order in sequence of the transform. If a pre-reg transform, it will not be reversed like a sequence of elastix transforms would to make the composite ITK transform

append(other) → None

Concatenate transformation sequences.

Parameters:

other (RegTransformSeq) – Append a RegTransformSeq to another

composed_linear_mats: Dict[str, ndarray] | None = None

property composite_transform: CompositeTransform

Composite ITK transform from transformation sequence

property n_transforms: int

Number of transformations in sequence.

property output_size: Tuple[int, int]

Output size of image resampled by transform, initially determined from the last transformation in the chain

property output_spacing: Tuple[float, float] | Tuple[int, int]

Output spacing of image resampled by transform, initially determined from the last transformation in the chain

reg_transforms: List[RegTransform] = []

reg_transforms_itk_order: List[RegTransform] = []

resampler: ResampleImageFilter | None = None

set_output_spacing(spacing: Tuple[float, float] | Tuple[int, int]) → None

Method that allows setting the output spacing of the resampler to resampled to any pixel spacing desired. This will also change the output_size to match.

Parameters:

spacing (tuple of float) – Spacing to set the new image. Will also change the output size to match.

transform_points(pt_data: ndarray, px_idx=True, source_res=1, output_idx=True) → ndarray

Transform point sets using the transformation chain :param pt_data: Point data in xy order :type pt_data: np.ndarray :param px_idx: Whether point data is in pixel or physical coordinate sapce :type px_idx: bool :param source_res: spacing of the pixels associated with pt_data if they are not in physical coordinate space :type source_res: float :param output_idx: return transformed points to pixel indices in the output_spacing’s reference space :type output_idx: bool

Returns:

tformed_pts – Transformed points

Return type:

np.ndarray

property transform_seq_idx: List[int]

Transformation sequence for all combined transformations.