Source code for ants.core.ants_transform
import numpy as np
__all__ = ['ANTsTransform',
'set_ants_transform_parameters',
'get_ants_transform_parameters',
'get_ants_transform_fixed_parameters',
'set_ants_transform_fixed_parameters',
'apply_ants_transform',
'apply_ants_transform_to_point',
'apply_ants_transform_to_vector',
'apply_ants_transform_to_image',
'invert_ants_transform',
'compose_ants_transforms',
'transform_index_to_physical_point',
'transform_physical_point_to_index']
from . import ants_image as iio
from .. import utils
[docs]class ANTsTransform(object):
def __init__(self, precision='float', dimension=3, transform_type='AffineTransform', pointer=None):
"""
NOTE: This class should never be initialized directly by the user.
Initialize an ANTsTransform object.
Arguments
---------
pixeltype : string
dimension : integer
transform_type : stirng
pointer : py::capsule (optional)
"""
self.precision = precision
self.dimension = dimension
self.transform_type = transform_type
self.type = transform_type
if pointer is None:
libfn = utils.get_lib_fn('newAntsTransform%s%i' % (utils.short_ptype(precision), dimension))
pointer = libfn(precision, dimension, transform_type)
self.pointer = pointer
# suffix to append to all c++ library function calls
self._libsuffix = '%s%i' % (utils.short_ptype(self.precision), self.dimension)
@property
def parameters(self):
""" Get parameters of transform """
libfn = utils.get_lib_fn('getTransformParameters%s'%self._libsuffix)
return np.asarray(libfn(self.pointer), order='F')#.reshape((self.dimension, self.dimension+1), order='F')
[docs] def set_parameters(self, parameters):
""" Set parameters of transform """
if not isinstance(parameters, np.ndarray):
parameters = np.asarray(parameters)
# if in two dimensions, flatten with fortran ordering
if parameters.ndim > 1:
parameters = parameters.flatten(order='F')
libfn = utils.get_lib_fn('setTransformParameters%s'%self._libsuffix)
libfn(self.pointer, parameters.tolist())
@property
def fixed_parameters(self):
""" Get parameters of transform """
libfn = utils.get_lib_fn('getTransformFixedParameters%s'%self._libsuffix)
return np.asarray(libfn(self.pointer))
[docs] def set_fixed_parameters(self, parameters):
""" Set parameters of transform """
if not isinstance(parameters, np.ndarray):
parameters = np.asarray(parameters)
libfn = utils.get_lib_fn('setTransformFixedParameters%s'%self._libsuffix)
libfn(self.pointer, parameters.tolist())
[docs] def invert(self):
""" Invert the transform """
libfn = utils.get_lib_fn('inverseTransform%s' % (self._libsuffix))
inv_tx_ptr = libfn(self.pointer)
new_tx = ANTsTransform(precision=self.precision, dimension=self.dimension,
transform_type=self.transform_type, pointer=inv_tx_ptr)
return new_tx
[docs] def apply(self, data, data_type='point', reference=None, **kwargs):
"""
Apply transform to data
"""
if data_type == 'point':
return self.apply_to_point(data)
elif data_type == 'vector':
return self.apply_to_vector(data)
elif data_type == 'image':
return self.apply_to_image(data, reference, **kwargs)
[docs] def apply_to_point(self, point):
"""
Apply transform to a point
Arguments
---------
point : list/tuple
point to which the transform will be applied
Returns
-------
list : transformed point
Example
-------
>>> import ants
>>> tx = ants.new_ants_transform()
>>> params = tx.parameters
>>> tx.set_parameters(params*2)
>>> pt2 = tx.apply_to_point((1,2,3)) # should be (2,4,6)
"""
libfn = utils.get_lib_fn('transformPoint%s' % (self._libsuffix))
return tuple(libfn(self.pointer, point))
[docs] def apply_to_vector(self, vector):
"""
Apply transform to a vector
Arguments
---------
vector : list/tuple
vector to which the transform will be applied
Returns
-------
list : transformed vector
"""
if isinstance(vector, np.ndarray):
vector = vector.tolist()
libfn = utils.get_lib_fn('transformVector%s' % (self._libsuffix))
return np.asarray(libfn(self.pointer, vector))
[docs] def apply_to_image(self, image, reference=None, interpolation='linear'):
"""
Apply transform to an image
Arguments
---------
image : ANTsImage
image to which the transform will be applied
reference : ANTsImage
target space for transforming image
interpolation : string
type of interpolation to use. Options are:
linear
nearestneighbor
multilabel
gaussian
bspline
cosinewindowedsinc
welchwindowedsinc
hammingwindoweddinc
lanczoswindowedsinc
genericlabel
Returns
-------
list : transformed vector
"""
if reference is None:
reference = image.clone()
interpolation = interpolation.lower()
tform_fn = utils.get_lib_fn('transformImage%s%s' % (self._libsuffix, image._libsuffix))
reference = reference.clone(image.pixeltype)
img_ptr = tform_fn(self.pointer, image.pointer, reference.pointer, interpolation)
return iio.ANTsImage(pixeltype=image.pixeltype,
dimension=image.dimension,
components=image.components,
pointer=img_ptr)
def __repr__(self):
s = "ANTsTransform\n" +\
'\t {:<10} : {}\n'.format('Type', self.type)+\
'\t {:<10} : {}\n'.format('Dimension', self.dimension)+\
'\t {:<10} : {}\n'.format('Precision', self.precision)
return s
# verbose functions for ANTsR compatibility
[docs]def set_ants_transform_parameters(transform, parameters):
"""
Set parameters of an ANTsTransform
ANTsR function: `setAntsrTransformParameters`
"""
transform.set_parameters(parameters)
[docs]def get_ants_transform_parameters(transform):
"""
Get parameters of an ANTsTransform
ANTsR function: `getAntsrTransformParameters`
"""
return transform.parameters
[docs]def set_ants_transform_fixed_parameters(transform, parameters):
"""
Set fixed parameters of an ANTsTransform
ANTsR function: `setAntsrTransformFixedParameters`
"""
transform.set_fixed_parameters(parameters)
[docs]def get_ants_transform_fixed_parameters(transform):
"""
Get fixed parameters of an ANTsTransform
ANTsR function: `getAntsrTransformFixedParameters`
"""
return transform.fixed_parameters
[docs]def apply_ants_transform(transform, data, data_type="point", reference=None, **kwargs):
"""
Apply ANTsTransform to data
ANTsR function: `applyAntsrTransform`
Arguments
---------
transform : ANTsTransform
transform to apply to image
data : ndarray/list/tuple
data to which transform will be applied
data_type : string
type of data
Options :
'point'
'vector'
'image'
reference : ANTsImage
target space for transforming image
kwargs : kwargs
additional options passed to `apply_ants_transform_to_image`
Returns
-------
ANTsImage if data_type == 'point'
OR
tuple if data_type == 'point' or data_type == 'vector'
"""
return transform.apply(data, data_type, reference, **kwargs)
[docs]def apply_ants_transform_to_point(transform, point):
"""
Apply transform to a point
ANTsR function: `applyAntsrTransformToPoint`
Arguments
---------
point : list/tuple
point to which the transform will be applied
Returns
-------
tuple : transformed point
Example
-------
>>> import ants
>>> tx = ants.new_ants_transform()
>>> params = tx.parameters
>>> tx.set_parameters(params*2)
>>> pt2 = tx.apply_to_point((1,2,3)) # should be (2,4,6)
"""
return transform.apply_to_point(point)
[docs]def apply_ants_transform_to_vector(transform, vector):
"""
Apply transform to a vector
ANTsR function: `applyAntsrTransformToVector`
Arguments
---------
vector : list/tuple
vector to which the transform will be applied
Returns
-------
tuple : transformed vector
"""
return transform.apply_to_vector(vector)
[docs]def apply_ants_transform_to_image(transform, image, reference, interpolation='linear'):
"""
Apply transform to an image
ANTsR function: `applyAntsrTransformToImage`
Arguments
---------
image : ANTsImage
image to which the transform will be applied
reference : ANTsImage
reference image
interpolation : string
type of interpolation to use. Options are:
linear
nearestneighbor
multilabel
gaussian
bspline
cosinewindowedsinc
welchwindowedsinc
hammingwindoweddinc
lanczoswindowedsinc
genericlabel
Returns
-------
list : transformed vector
Example
-------
>>> import ants
>>> img = ants.image_read(ants.get_ants_data("r16")).clone('float')
>>> tx = ants.new_ants_transform(dimension=2)
>>> tx.set_parameters((0.9,0,0,1.1,10,11))
>>> img2 = tx.apply_to_image(img, img)
"""
return transform.apply_to_image(image, reference, interpolation)
[docs]def invert_ants_transform(transform):
"""
Invert ANTsTransform
ANTsR function: `invertAntsrTransform`
Example
-------
>>> import ants
>>> img = ants.image_read(ants.get_ants_data("r16")).clone('float')
>>> tx = ants.new_ants_transform(dimension=2)
>>> tx.set_parameters((0.9,0,0,1.1,10,11))
>>> img_transformed = tx.apply_to_image(img, img)
>>> inv_tx = tx.invert()
>>> img_orig = inv_tx.apply_to_image(img_transformed, img_transformed)
"""
return transform.invert()
[docs]def compose_ants_transforms(transform_list):
"""
Compose multiple ANTsTransform's together
ANTsR function: `composeAntsrTransforms`
Arguments
---------
transform_list : list/tuple of ANTsTransform object
list of transforms to compose together
Returns
-------
ANTsTransform
one transform that contains all given transforms
Example
-------
>>> import ants
>>> img = ants.image_read(ants.get_ants_data("r16")).clone('float')
>>> tx = ants.new_ants_transform(dimension=2)
>>> tx.set_parameters((0.9,0,0,1.1,10,11))
>>> inv_tx = tx.invert()
>>> single_tx = ants.compose_ants_transforms([tx, inv_tx])
>>> img_orig = single_tx.apply_to_image(img, img)
>>> rRotGenerator = ants.contrib.RandomRotate2D( ( 0, 40 ), reference=img )
>>> rShearGenerator=ants.contrib.RandomShear2D( (0,50), reference=img )
>>> tx1 = rRotGenerator.transform()
>>> tx2 = rShearGenerator.transform()
>>> rSrR = ants.compose_ants_transforms([tx1, tx2])
>>> rSrR.apply_to_image( img )
"""
precision = transform_list[0].precision
dimension = transform_list[0].dimension
for tx in transform_list:
if precision != tx.precision:
raise ValueError('All transforms must have the same precision')
if dimension != tx.dimension:
raise ValueError('All transforms must have the same dimension')
tx_ptr_list = list(reversed([tf.pointer for tf in transform_list]))
libfn = utils.get_lib_fn('composeTransforms%s' % (transform_list[0]._libsuffix))
itk_composed_tx = libfn(tx_ptr_list, precision, dimension)
return ANTsTransform(precision=precision, dimension=dimension,
transform_type='CompositeTransform', pointer=itk_composed_tx)
[docs]def transform_index_to_physical_point(image, index):
"""
Get spatial point from index of an image.
ANTsR function: `antsTransformIndexToPhysicalPoint`
Arguments
---------
img : ANTsImage
image to get values from
index : list or tuple or numpy.ndarray
location in image
Returns
-------
tuple
Example
-------
>>> import ants
>>> import numpy as np
>>> img = ants.make_image((10,10),np.random.randn(100))
>>> pt = ants.transform_index_to_physical_point(img, (2,2))
"""
if not isinstance(image, iio.ANTsImage):
raise ValueError('image must be ANTsImage type')
if isinstance(index, np.ndarray):
index = index.tolist()
if not isinstance(index, (tuple,list)):
raise ValueError('index must be tuple or list')
if len(index) != image.dimension:
raise ValueError('len(index) != image.dimension')
index = [i+1 for i in index]
ndim = image.dimension
ptype = image.pixeltype
libfn = utils.get_lib_fn('TransformIndexToPhysicalPoint%s%i' % (utils.short_ptype(ptype), ndim))
point = libfn(image.pointer, [list(index)])
return np.array(point[0])
[docs]def transform_physical_point_to_index(image, point):
"""
Get index from spatial point of an image.
ANTsR function: `antsTransformPhysicalPointToIndex`
Arguments
---------
image : ANTsImage
image to get values from
point : list or tuple or numpy.ndarray
point in image
Returns
-------
tuple
Example
-------
>>> import ants
>>> import numpy as np
>>> img = ants.make_image((10,10),np.random.randn(100))
>>> idx = ants.transform_physical_point_to_index(img, (2,2))
>>> img.set_spacing((2,2))
>>> idx2 = ants.transform_physical_point_to_index(img, (4,4))
"""
if not isinstance(image, iio.ANTsImage):
raise ValueError('image must be ANTsImage type')
if isinstance(point, np.ndarray):
point = point.tolist()
if not isinstance(point, (list,tuple)):
raise ValueError('point must be list, tuple, or np.ndarray')
if len(point) != image.dimension:
raise ValueError('len(index) != image.dimension')
ndim = image.dimension
ptype = image.pixeltype
libfn = utils.get_lib_fn('TransformPhysicalPointToIndex%s%i'%(utils.short_ptype(ptype),ndim))
index = libfn(image.pointer, [list(point)])
index = [i-1 for i in index[0]]
return np.array(index)
