def image_postprocessing(self, image):
"""
Processes input image files. Blurs, crops, pads, and normalizes according to the parameters.
:param image: The sitk image.
:return: The processed sitk image.
"""
if self.image_gaussian_blur_sigma > 0:
image = gaussian(image, [self.image_gaussian_blur_sigma, self.image_gaussian_blur_sigma])
image_float = sitk.Cast(image, sitk.sitkFloat32)
image_float = normalize_robust_sitk(image_float, (-1, 1), consideration_factors=self.normalization_consideration_factors)
return image_float
def data_sources(self,
iterator,
cached,
use_landmarks=True,
image_extension='.nii.gz'):
"""
Returns the data sources that load data.
{
'image_datasource:' ImageDataSource that loads the image files.
'landmarks_datasource:' LandmarkDataSource that loads the landmark coordinates.
}
:param iterator: The iterator node object.
:param cached: If True, use CachedImageDataSource instead of ImageDataSource.
:param use_landmarks: If True, create a landmarks datasource.
:param image_extension: The extension of the image files.
:return: A dict of data sources.
"""
if self.smoothing_sigma is not None and self.smoothing_sigma > 0:
preprocessing = lambda x: sitk.Cast(
gaussian(x, self.smoothing_sigma), sitk.sitkInt16)
else:
preprocessing = None
if cached:
image_datasource = CachedImageDataSource(
self.image_base_folder,
'',
'',
image_extension,
preprocessing=preprocessing,
cache_maxsize=25000,
name='image_datasource',
parents=[iterator])
else:
image_datasource = ImageDataSource(self.image_base_folder,
'',
'',
image_extension,
preprocessing=preprocessing,
name='image_datasource',
parents=[iterator])
data_sources_dict = {}
data_sources_dict['image_datasource'] = image_datasource
if use_landmarks:
landmark_datasource = LandmarkDataSource(
self.point_list_file_name,
self.num_landmarks,
self.dim,
name='landmarks_datasource',
parents=[iterator])
data_sources_dict['landmarks_datasource'] = landmark_datasource
return data_sources_dict
def image_preprocessing(self, image):
if self.image_gaussian_blur_sigma > 0:
image = gaussian(image, self.image_gaussian_blur_sigma)
if self.crop_image_size is not None:
image = sitk.Crop(image, self.crop_image_size,
self.crop_image_size)
image.SetOrigin([0] * image.GetDimension())
if self.pad_image:
pad_size = image.GetSize()
pad_size = [int(s / 2) for s in pad_size]
image = sitk.MirrorPad(image, pad_size, pad_size)
image_float = sitk.Cast(image, sitk.sitkFloat32)
image_float = normalize_robust_sitk(
image_float, (-1, 1),
consideration_factors=self.normalization_consideration_factors)
return image_float
def image_postprocessing(self, image):
"""
Processes input image files. Blurs, crops, pads, and normalizes according to the parameters.
:param image: The sitk image.
:return: The processed sitk image.
"""
if self.image_gaussian_blur_sigma > 0:
image = gaussian(image, [self.image_gaussian_blur_sigma, self.image_gaussian_blur_sigma])
if self.crop_image_size is not None:
image = sitk.Crop(image, self.crop_image_size, self.crop_image_size)
if self.pad_image:
pad_size = image.GetSize()
pad_size = [int(s / 2) for s in pad_size]
image = sitk.MirrorPad(image, pad_size, pad_size)
image_float = sitk.Cast(image, sitk.sitkFloat32)
image_float = normalize_robust_sitk(image_float, (-1, 1), consideration_factors=self.normalization_consideration_factors)
return image_float
def image_preprocessing(self, image):
if self.image_gaussian_sigma > 0.0:
return gaussian(image, self.image_gaussian_sigma)
else:
return image