def determine_whether_to_use_mask_for_norm(self):
# only use the nonzero mask for normalization of the cropping based on it resulted in a decrease in
# image size (this is an indication that the data is something like brats/isles and then we want to
# normalize in the brain region only)
modalities = self.dataset_properties['modalities']
num_modalities = len(list(modalities.keys()))
use_nonzero_mask_for_norm = OrderedDict()
for i in range(num_modalities):
if "CT" in modalities[i]:
use_nonzero_mask_for_norm[i] = False
else:
all_size_reductions = []
for k in self.dataset_properties['size_reductions'].keys():
all_size_reductions.append(
self.dataset_properties['size_reductions'][k])
if np.median(all_size_reductions) < 3 / 4.:
print("using nonzero mask for normalization")
use_nonzero_mask_for_norm[i] = True
else:
print("not using nonzero mask for normalization")
use_nonzero_mask_for_norm[i] = False
for c in self.list_of_cropped_npz_files:
case_identifier = get_case_identifier_from_npz(c)
properties = self.load_properties_of_cropped(case_identifier)
properties['use_nonzero_mask_for_norm'] = use_nonzero_mask_for_norm
self.save_properties_of_cropped(case_identifier, properties)
use_nonzero_mask_for_normalization = use_nonzero_mask_for_norm
return use_nonzero_mask_for_normalization
def write_normalization_scheme_to_patients(self):
"""
This is used for test set preprocessing
:return:
"""
for c in self.list_of_cropped_npz_files:
case_identifier = get_case_identifier_from_npz(c)
properties = self.load_properties_of_cropped(case_identifier)
properties['use_nonzero_mask_for_norm'] = self.plans['use_mask_for_norm']
self.save_properties_of_cropped(case_identifier, properties)
