def _resize_image_test(self, image, target_shape):
original_image_shape = image.shape
new_image = resize(image, target_shape)
self.assertEqual(new_image.shape, target_shape)
new_image = resize(new_image,
original_image_shape,
interpolation="linear")
self.assertEqual(new_image.shape, original_image_shape)
def restore_dimension(image_label, slices, affine):
old_shape = tuple([x.stop - x.start for x in slices])
old_cropped_image = resize(image_label, old_shape, interpolation='nearest')
rotated_image = np.rot90(old_cropped_image.get_data(), 2)
result = np.zeros(original_shape, dtype=np.uint8)
tp_slices = tuple(slices)
result[tp_slices] = rotated_image
return nib.Nifti1Image(result, affine)
def find_downsized_info(training_data_files, input_shape):
foreground = get_complete_foreground(training_data_files)
crop_slices = crop_img(foreground, return_slices=True, copy=True)
cropped = crop_img_to(foreground, crop_slices, copy=True)
final_image = resize(cropped,
new_shape=input_shape,
interpolation="nearest")
return crop_slices, final_image.affine, final_image.header
def test_images_align():
data = np.arange(1, 9).reshape((2, 2, 2))
affine = np.diag(np.ones(4) * 2)
affine[3, 3] = 1
image_nib = nib.Nifti1Image(data, affine=affine)
new_image_nib = resize(image_nib, (4, 4, 4), interpolation="nearest")
assert np.all(new_image_nib.get_data()[0] == np.asarray(
[[1, 1, 2, 2], [1, 1, 2, 2], [3, 3, 4, 4], [3, 3, 4, 4]]))
assert np.all(new_image_nib.affine == np.asarray(
[[1., 0., 0., -0.5], [0., 1., 0., -0.5], [0., 0., 1., -0.5],
[0., 0., 0., 1.]]))
def crop_subject_modals(subject_modal_imgs, input_shape, slices):
subject_data = []
affine = None
for i, modal_img in enumerate(subject_modal_imgs):
modal_img = fix_shape(modal_img)
modal_img = crop_img_to(modal_img, slices, copy=True)
new_img = resize(modal_img,
new_shape=input_shape,
interpolation='linear')
subject_data.append(new_img.get_data())
if i == 0:
affine = new_img.get_affine()
subject_data = np.asarray(subject_data)
return subject_data, affine
def resize_modal_image(subject_modal_imgs,
target_shape,
interpolation='linear'):
subject_data = []
affine = None
for i, modal_img in enumerate(subject_modal_imgs):
modal_img = fix_shape(modal_img)
new_img = resize(modal_img,
new_shape=target_shape,
interpolation=interpolation)
subject_data.append(new_img.get_data())
if i == 0:
affine = new_img.get_affine()
subject_data = np.asarray(subject_data)
return subject_data, affine
def segmentation_for_patient(subject_fd,
config,
output_path,
model=None,
mode='size_same_input'):
if model is None:
model = load_old_model(config)
subject_name = os.path.basename(subject_fd)
image_mris, original_affine, foreground = get_subject_tensor(
subject_fd, subject_name)
if mode == 'size_same_input':
slices = get_slices(foreground)
subject_data_fixed_size, affine = crop_subject_modals(
image_mris, input_shape, slices)
elif mode == 'size_interpolate':
target_shape = tuple(config['inference_shape'])
subject_data_fixed_size, affine = resize_modal_image(
image_mris, target_shape)
else:
print('Do not support mode {} for inference'.format(mode))
return
subject_tensor = normalize_data(subject_data_fixed_size)
subject_tensor = np.expand_dims(subject_tensor, axis=0)
output_predict = predict(model, subject_tensor, affine)
if mode == 'size_same_input':
output = restore_dimension(output_predict, slices, original_affine)
elif mode == 'size_interpolate':
output = resize(output_predict,
new_shape=original_shape,
interpolation='nearest')
else:
print('Do not support mode {} for inference'.format(mode))
return
output_fd = os.path.join(output_path, subject_name)
if not os.path.exists(output_fd):
os.makedirs(output_fd)
output_file = os.path.join(
output_fd, '{}_prediction{}'.format(subject_name, extension))
output.to_filename(output_file)
print('Patient {} is done !'.format(subject_fd))
def _resize_image_test(image, target_shape):
original_image_shape = image.shape
new_image = resize(image, target_shape)
assert new_image.shape == target_shape
new_image = resize(new_image, original_image_shape, interpolation="linear")
assert new_image.shape == original_image_shape
