def postprocessing_random(self, image):
"""
Performs random augmentations of a grayscale image. Augmentation consists of random gamma correction,
random intensity shift/scale per video and per frame.
:param image: The grayscale image to augment.
:return: The augmented grayscale image.
"""
random_lambda = float_uniform(0.6, 1.4)
image = change_gamma_unnormalized(image, random_lambda)
image = ShiftScaleClamp(random_shift=0.65, random_scale=0.65)(image)
if len(image.shape) == 4:
for i in range(image.shape[self.video_frame_stack_axis]):
current_slice = [slice(None), slice(None)]
current_slice.insert(self.video_frame_stack_axis, slice(i, i + 1))
image[tuple(current_slice)] = ShiftScaleClamp(random_shift=0.1, random_scale=0.1)(image[tuple(current_slice)])
return image
def intensity_postprocessing_mr(self, image):
"""
Intensity postprocessing for MR input.
:param image: The np input image.
:return: The processed image.
"""
image = normalize_robust(image)
return ShiftScaleClamp(clamp_min=-1.0)(image)
def binary_labels(self, image):
"""
Converts an instance label image into a binary label. All instances will be set to 1.
:param image: The instance label image.
:return: A list of np arrays. First, is the background. Second is the foreground.
"""
all_seg = ShiftScaleClamp(clamp_min=0, clamp_max=1)(image)
return split_label_image(all_seg, [0, 1])
def intensity_postprocessing_augmented(self, image):
"""
Intensity postprocessing. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
normalized = normalize_robust(image)
return ShiftScaleClamp(random_shift=0.15,
random_scale=0.15)(normalized)
def intensity_postprocessing_mr_random(self, image):
"""
Intensity postprocessing for MR input. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
image = normalize_robust(image)
return ShiftScaleClamp(random_shift=0.2,
random_scale=0.4,
clamp_min=-1.0)(image)
def intensity_postprocessing(self, image):
"""
Intensity postprocessing.
:param image: The np input image.
:return: The processed image.
"""
return ShiftScaleClamp(shift=0,
scale=1 / 2048,
clamp_min=-1.0,
clamp_max=1.0)(image)
def intensity_postprocessing_mr_random(self, image):
"""
Intensity postprocessing for MR input. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
image = change_gamma_unnormalized(image, float_uniform(0.5, 1.5))
image = normalize_robust(image, consideration_factors=(0.1, 0.1))
return ShiftScaleClamp(random_shift=0.6,
random_scale=0.6,
clamp_min=-1.0)(image)
def intensity_postprocessing_augmented(self, image):
"""
Intensity postprocessing. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
return ShiftScaleClamp(shift=0,
scale=1 / 2048,
clamp_min=-1.0,
clamp_max=1.0,
random_shift=0.15,
random_scale=0.15)(image)
def intensity_postprocessing_ct_random(self, image):
"""
Intensity postprocessing for CT input. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
return ShiftScaleClamp(shift=0,
scale=1 / 2048,
random_shift=self.random_intensity_shift,
random_scale=self.random_intensity_scale,
clamp_min=-1.0,
clamp_max=1.0)(image)
def intensity_postprocessing_ct(self, image):
"""
Intensity postprocessing for CT input.
:param image: The np input image.
:return: The processed image.
"""
if not self.normalize_zero_mean_unit_variance:
output = ShiftScaleClamp(shift=0,
scale=1 / 2048,
clamp_min=-1.0,
clamp_max=1.0)(image)
else:
output = normalize_zero_mean_unit_variance(image)
return output
def intensity_postprocessing_ct_random(self, image):
"""
Intensity postprocessing for CT input. Random augmentation version.
:param image: The np input image.
:return: The processed image.
"""
if not self.normalize_zero_mean_unit_variance:
random_lambda = float_uniform(0.9, 1.1)
image = change_gamma_unnormalized(image, random_lambda)
output = ShiftScaleClamp(shift=0,
scale=1 / 2048,
random_shift=self.random_intensity_shift,
random_scale=self.random_intensity_scale,
clamp_min=-1.0,
clamp_max=1.0)(image)
else:
random_lambda = float_uniform(0.9, 1.1)
image = change_gamma_unnormalized(image, random_lambda)
output = normalize_zero_mean_unit_variance(image)
return output
def test(self):
print('Testing...')
if self.data_format == 'channels_first':
np_channel_index = 0
else:
np_channel_index = 3
heatmap_maxima = HeatmapTest(np_channel_index, False)
landmark_statistics = LandmarkStatistics()
landmarks = {}
for i in range(self.dataset_val.num_entries()):
dataset_entry = self.dataset_val.get_next()
current_id = dataset_entry['id']['image_id']
datasources = dataset_entry['datasources']
image_datasource = datasources['image_datasource']
landmarks_datasource = datasources['landmarks_datasource']
if not self.cropped_training:
image, heatmaps, heatmap_transform = self.test_full_image(
dataset_entry)
else:
image, heatmaps, heatmap_transform = self.test_cropped_image(
dataset_entry)
utils.io.image.write_np(
ShiftScaleClamp(scale=255, clamp_min=0,
clamp_max=255)(heatmaps).astype(np.uint8),
self.output_file_for_current_iteration(current_id +
'_heatmaps.mha'))
utils.io.image.write_np(
image,
self.output_file_for_current_iteration(current_id +
'_image.mha'))
predicted_landmarks = heatmap_maxima.get_landmarks(
heatmaps, image_datasource, self.image_spacing,
heatmap_transform)
landmarks[current_id] = predicted_landmarks
landmark_statistics.add_landmarks(current_id, predicted_landmarks,
landmarks_datasource)
tensorflow_train.utils.tensorflow_util.print_progress_bar(
i,
self.dataset_val.num_entries(),
prefix='Testing ',
suffix=' complete')
tensorflow_train.utils.tensorflow_util.print_progress_bar(
self.dataset_val.num_entries(),
self.dataset_val.num_entries(),
prefix='Testing ',
suffix=' complete')
print(landmark_statistics.get_pe_overview_string())
print(landmark_statistics.get_correct_id_string(20.0))
summary_values = OrderedDict(
zip(
self.point_statistics_names,
list(landmark_statistics.get_pe_statistics()) +
[landmark_statistics.get_correct_id(20)]))
# finalize loss values
self.val_loss_aggregator.finalize(self.current_iter, summary_values)
utils.io.landmark.save_points_csv(
landmarks, self.output_file_for_current_iteration('points.csv'))
overview_string = landmark_statistics.get_overview_string(
[2, 2.5, 3, 4, 10, 20], 10, 20.0)
utils.io.text.save_string_txt(
overview_string,
self.output_file_for_current_iteration('eval.txt'))
