def run_main(context):
no_transform = torch_transforms.Compose([
imed_transforms.CenterCrop([128, 128]),
imed_transforms.NumpyToTensor(),
imed_transforms.NormalizeInstance(),
])
out_dir = context["path_output"]
split_dct = joblib.load(os.path.join(out_dir, "split_datasets.joblib"))
metadata_dct = {}
for subset in ['train', 'valid', 'test']:
metadata_dct[subset] = {}
ds = imed_loader.BidsDataset(context["path_data"],
subject_lst=split_dct[subset],
contrast_lst=context["contrast_train_validation"]
if subset != "test" else context["contrast_test"],
transform=no_transform,
slice_filter_fn=imed_loader_utils.SliceFilter())
for m in metadata_type:
if m in metadata_dct:
metadata_dct[subset][m] = [v for m_lst in [metadata_dct[subset][m], ds.metadata[m]] for v in m_lst]
else:
metadata_dct[subset][m] = ds.metadata[m]
cluster_dct = joblib.load(os.path.join(out_dir, "clustering_models.joblib"))
out_dir = os.path.join(out_dir, "cluster_metadata")
if not os.path.isdir(out_dir):
os.makedirs(out_dir)
for m in metadata_type:
values = [v for s in ['train', 'valid', 'test'] for v in metadata_dct[s][m]]
print('\n{}: Min={}, Max={}, Median={}'.format(m, min(values), max(values), np.median(values)))
plot_decision_boundaries(metadata_dct, cluster_dct[m], metadata_range[m], m, os.path.join(out_dir, m + '.png'))
def run_main(args):
with open(args.c, "r") as fhandle:
context = json.load(fhandle)
transform_lst = torch_transforms.Compose([
imed_transforms.Resample(wspace=0.75, hspace=0.75),
imed_transforms.CenterCrop([128, 128]),
imed_transforms.NumpyToTensor(),
imed_transforms.NormalizeInstance(),
])
train_lst, valid_lst, test_lst = imed_loader_utils.split_dataset(
context["bids_path"], context["center_test"], context["split_method"],
context["random_seed"])
balance_dct = {}
for ds_lst, ds_name in zip([train_lst, valid_lst, test_lst],
['train', 'valid', 'test']):
print("\nLoading {} set.\n".format(ds_name))
ds = imed_loader.BidsDataset(
context["bids_path"],
subject_lst=ds_lst,
target_suffix=context["target_suffix"],
contrast_lst=context["contrast_test"]
if ds_name == 'test' else context["contrast_train_validation"],
metadata_choice=context["metadata"],
contrast_balance=context["contrast_balance"],
transform=transform_lst,
slice_filter_fn=imed_utils.SliceFilter())
print("Loaded {} axial slices for the {} set.".format(
len(ds), ds_name))
ds_loader = DataLoader(ds,
batch_size=1,
shuffle=False,
pin_memory=False,
collate_fn=imed_loader_utils.imed_collate,
num_workers=1)
balance_lst = []
for i, batch in enumerate(ds_loader):
gt_sample = batch["gt"].numpy().astype(np.int)[0, 0, :, :]
nb_ones = (gt_sample == 1).sum()
nb_voxels = gt_sample.size
balance_lst.append(nb_ones * 100.0 / nb_voxels)
balance_dct[ds_name] = balance_lst
for ds_name in balance_dct:
print('\nClass balance in {} set:'.format(ds_name))
print_stats(balance_dct[ds_name])
print('\nClass balance in full set:')
print_stats([e for d in balance_dct for e in balance_dct[d]])
def run_main(context):
no_transform = torch_transforms.Compose([
imed_transforms.CenterCrop([128, 128]),
imed_transforms.NumpyToTensor(),
imed_transforms.NormalizeInstance(),
])
out_dir = context["log_directory"]
metadata_dct = {}
for subset in ['train', 'validation', 'test']:
metadata_dct[subset] = {}
for bids_ds in tqdm(context["bids_path_" + subset],
desc="Loading " + subset + " set"):
ds = imed_loader.BidsDataset(
bids_ds,
contrast_lst=context["contrast_train_validation"]
if subset != "test" else context["contrast_test"],
transform=no_transform,
slice_filter_fn=imed_loader_utils.SliceFilter())
for m in metadata_type:
if m in metadata_dct:
metadata_dct[subset][m] = [
v
for m_lst in [metadata_dct[subset][m], ds.metadata[m]]
for v in m_lst
]
else:
metadata_dct[subset][m] = ds.metadata[m]
for m in metadata_type:
metadata_dct[subset][m] = list(set(metadata_dct[subset][m]))
with open(out_dir + "/metadata_config.json", 'w') as fp:
json.dump(metadata_dct, fp)
return