def get_pair_data(self):
"""Return the tuple (input, ground truth) with the data content in numpy array."""
cache_mode = 'fill' if self.cache else 'unchanged'
input_data = []
for handle in self.input_handle:
hwd_oriented = imed_loader_utils.orient_img_hwd(
handle.get_fdata(cache_mode, dtype=np.float32),
self.slice_axis)
input_data.append(hwd_oriented)
gt_data = []
# Handle unlabeled data
if self.gt_handle is None:
gt_data = None
for gt in self.gt_handle:
if gt is not None:
hwd_oriented = imed_loader_utils.orient_img_hwd(
gt.get_fdata(cache_mode, dtype=np.float32),
self.slice_axis)
data_type = np.float32 if self.soft_gt else np.uint8
gt_data.append(hwd_oriented.astype(data_type))
else:
gt_data.append(
np.zeros(imed_loader_utils.orient_shapes_hwd(
self.input_handle[0].shape, self.slice_axis),
dtype=np.float32).astype(np.uint8))
return input_data, gt_data
def get_pair_data(self):
"""Return the tuple (input, ground truth) with the data content in numpy array."""
cache_mode = 'fill' if self.cache else 'unchanged'
input_data = []
for handle in self.input_handle:
hwd_oriented = imed_loader_utils.orient_img_hwd(handle.get_fdata(cache_mode, dtype=np.float32),
self.slice_axis)
input_data.append(hwd_oriented)
gt_data = []
# Handle unlabeled data
if self.gt_handle is None:
gt_data = None
for gt in self.gt_handle:
if gt is not None:
if not isinstance(gt, list): # this tissue has annotation from only one rater
hwd_oriented = imed_loader_utils.orient_img_hwd(gt.get_fdata(cache_mode, dtype=np.float32),
self.slice_axis)
gt_data.append(hwd_oriented)
else: # this tissue has annotation from several raters
hwd_oriented_list = [
imed_loader_utils.orient_img_hwd(gt_rater.get_fdata(cache_mode, dtype=np.float32),
self.slice_axis) for gt_rater in gt]
gt_data.append([hwd_oriented.astype(data_type) for hwd_oriented in hwd_oriented_list])
else:
gt_data.append(
np.zeros(imed_loader_utils.orient_shapes_hwd(self.input_handle[0].shape, self.slice_axis),
dtype=np.float32).astype(np.uint8))
return input_data, gt_data
def generate_bounding_box_file(subject_list,
model_path,
path_output,
gpu_id=0,
slice_axis=0,
contrast_lst=None,
keep_largest_only=True,
safety_factor=None):
"""Creates json file containing the bounding box dimension for each images. The file has the following format:
{"path/to/img.nii.gz": [[x1_min, x1_max, y1_min, y1_max, z1_min, z1_max],
[x2_min, x2_max, y2_min, y2_max, z2_min, z2_max]]}
where each list represents the coordinates of an object on the image (2 instance of a given object in this example).
Args:
subject_list (list): List of all subjects in the BIDS directory.
model_path (string): Path to object detection model.
path_output (string): Output path.
gpu_id (int): If available, GPU number.
slice_axis (int): Slice axis (0: sagittal, 1: coronal, 2: axial).
contrast_lst (list): Contrasts.
keep_largest_only (bool): Boolean representing if only the largest object of the prediction is kept.
safety_factor (list or tuple): Factors to multiply each dimension of the bounding box.
Returns:
dict: Dictionary containing bounding boxes related to their image.
"""
bounding_box_dict = {}
if safety_factor is None:
safety_factor = [1.0, 1.0, 1.0]
for subject in subject_list:
if subject.record["modality"] in contrast_lst:
subject_path = str(subject.record["absolute_path"])
object_mask, _ = imed_inference.segment_volume(model_path,
[subject_path],
gpu_id=gpu_id)
object_mask = object_mask[0]
if keep_largest_only:
object_mask = imed_postpro.keep_largest_object(object_mask)
mask_path = os.path.join(path_output, "detection_mask")
if not os.path.exists(mask_path):
os.mkdir(mask_path)
nib.save(object_mask,
os.path.join(mask_path,
subject_path.split("/")[-1]))
ras_orientation = nib.as_closest_canonical(object_mask)
hwd_orientation = imed_loader_utils.orient_img_hwd(
ras_orientation.get_fdata(), slice_axis)
bounding_boxes = get_bounding_boxes(hwd_orientation)
bounding_box_dict[subject_path] = [
adjust_bb_size(bb, safety_factor) for bb in bounding_boxes
]
file_path = os.path.join(path_output, 'bounding_boxes.json')
with open(file_path, 'w') as fp:
json.dump(bounding_box_dict, fp, indent=4)
return bounding_box_dict
def get_data(fname_in, axis):
"""Get data from fname along an axis.
Args:
fname_in string: image fname
axis int:
Returns:
nibabel, ndarray
"""
# Load image
input_img = nib.load(fname_in)
# Reorient as canonical
input_img = nib.as_closest_canonical(input_img)
# Get input data
input_data = input_img.get_fdata(dtype=np.float32)
# Reorient data
input_data = imed_loader_utils.orient_img_hwd(input_data, slice_axis=axis)
return input_img, input_data
def bounding_box_prior(fname_mask, metadata, slice_axis):
"""
Computes prior steps to a model requiring bounding box crop. This includes loading a mask of the ROI, orienting the
given mask into the following dimensions: (height, width, depth), extracting the bounding boxes and storing the
information in the metadata.
Args:
fname_mask (str): Filename containing the mask of the ROI
metadata (dict): Dictionary containing the image metadata
slice_axis (int): Slice axis (0: sagittal, 1: coronal, 2: axial)
"""
nib_prior = nib.load(fname_mask)
# orient image into HWD convention
nib_ras = nib.as_closest_canonical(nib_prior)
np_mask = nib_ras.get_fdata()[..., 0] if len(nib_ras.get_fdata().shape) == 4 else nib_ras.get_fdata()
np_mask = imed_loader_utils.orient_img_hwd(np_mask, slice_axis)
bounding_box = get_bounding_boxes(np_mask)
metadata['bounding_box'] = bounding_box[0]
def bounding_box_prior(fname_mask, metadata, slice_axis, safety_factor=None):
"""
Computes prior steps to a model requiring bounding box crop. This includes loading a mask of the ROI, orienting the
given mask into the following dimensions: (height, width, depth), extracting the bounding boxes and storing the
information in the metadata.
Args:
fname_mask (str): Filename containing the mask of the ROI
metadata (dict): Dictionary containing the image metadata
slice_axis (int): Slice axis (0: sagittal, 1: coronal, 2: axial)
safety_factor (list or tuple): Factors to multiply each dimension of the bounding box.
"""
nib_prior = nib.load(fname_mask)
# orient image into HWD convention
nib_ras = nib.as_closest_canonical(nib_prior)
np_mask = nib_ras.get_fdata()[..., 0] if len(
nib_ras.get_fdata().shape) == 4 else nib_ras.get_fdata()
np_mask = imed_loader_utils.orient_img_hwd(np_mask, slice_axis)
# Extract the bounding box from the list
bounding_box = get_bounding_boxes(np_mask)[0]
if safety_factor:
bounding_box = adjust_bb_size(bounding_box, safety_factor)
metadata[MetadataKW.BOUNDING_BOX] = bounding_box
