def test_write_5d(self):
out_dir = tempfile.mkdtemp()
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(12).reshape((1, 1, 3, 2, 2))
write_nifti(img,
image_name,
affine=np.diag([1]),
target_affine=np.diag([1.4]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0.0, 1.0], [2.0, 3.0]], [[4.0, 5.0], [6.0, 7.0]],
[[8.0, 9.0], [10.0, 11.0]]]]]),
)
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
image_name = os.path.join(out_dir, "test1.nii.gz")
img = np.arange(10).reshape((1, 1, 5, 1, 2))
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1, 3, 3]),
target_affine=np.diag([1.4, 2.0, 2, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0.0, 1.0]], [[4.0, 5.0]], [[8.0, 9.0]]]]]))
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]))
shutil.rmtree(out_dir)
def test_orientation(self, array, affine, reader_param, expected):
test_image = make_nifti_image(array, affine)
# read test cases
loader = LoadNifti(**reader_param)
load_result = loader(test_image)
if isinstance(load_result, tuple):
data_array, header = load_result
else:
data_array = load_result
header = None
if os.path.exists(test_image):
os.remove(test_image)
# write test cases
if header is not None:
write_nifti(data_array, test_image, header["affine"],
header.get("original_affine", None))
elif affine is not None:
write_nifti(data_array, test_image, affine)
saved = nib.load(test_image)
saved_affine = saved.affine
saved_data = saved.get_fdata()
if os.path.exists(test_image):
os.remove(test_image)
if affine is not None:
np.testing.assert_allclose(saved_affine, affine)
np.testing.assert_allclose(saved_data, expected)
def test_write_5d(self):
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, "test.nii.gz")
for p in TEST_NDARRAYS:
img = p(np.arange(12).reshape((1, 1, 3, 2, 2)))
write_nifti(img,
image_name,
affine=np.diag([1]),
target_affine=np.diag([1.4]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0.0, 1.0], [2.0, 3.0]],
[[4.0, 5.0], [6.0, 7.0]],
[[8.0, 9.0], [10.0, 11.0]]]]]),
)
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
image_name = os.path.join(out_dir, "test1.nii.gz")
img = p(np.arange(10).reshape((1, 1, 5, 1, 2)))
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1, 3, 3]),
target_affine=np.diag([1.4, 2.0, 2, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0.0, 1.0]], [[4.0, 5.0]], [[8.0, 9.0]]]]]))
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]))
def test_write_5d(self):
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, 'test.nii.gz')
img = np.arange(12).reshape((1, 1, 3, 2, 2))
write_nifti(img,
image_name,
affine=np.diag([1]),
target_affine=np.diag([1.4]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0., 1.], [2., 3.]], [[4., 5.], [6., 7.]],
[[8., 9.], [10., 11.]]]]]))
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, 'test.nii.gz')
img = np.arange(10).reshape((1, 1, 5, 1, 2))
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1, 3, 3]),
target_affine=np.diag([1.4, 2., 2, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(
out.get_fdata(),
np.array([[[[[0., 1.]], [[4., 5.]], [[8., 9.]]]]]))
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]))
def __call__(self, data):
logger.setLevel(data.get("logging", "INFO").upper())
ext = file_ext(data.get("image_path"))
dtype = data.get(self.key_dtype, None)
compress = data.get(self.key_compress, False)
write_to_file = data.get(self.key_write_to_file, True)
ext = data.get(self.key_extension) if data.get(self.key_extension) else ext
logger.info(f"Result ext: {ext}; write_to_file: {write_to_file}; dtype: {dtype}")
image_np = data[self.label]
meta_dict = data.get(f"{self.ref_image}_{self.meta_key_postfix}")
affine = meta_dict.get("affine") if meta_dict else None
logger.debug(f"Image: {image_np.shape}; Data Image: {data[self.label].shape}")
output_file = None
output_json = data.get(self.json, {})
if write_to_file:
output_file = tempfile.NamedTemporaryFile(suffix=ext).name
logger.debug(f"Saving Image to: {output_file}")
# Issue with slicer:: https://discourse.itk.org/t/saving-non-orthogonal-volume-in-nifti-format/2760/22
if self.nibabel and ext.lower() in [".nii", ".nii.gz"]:
logger.debug("Using MONAI write_nifti...")
write_nifti(image_np, output_file, affine=affine, output_dtype=dtype)
else:
write_itk(image_np, output_file, affine, dtype, compress)
return output_file, output_json
def test_consistency(self):
np.set_printoptions(suppress=True, precision=3)
test_image = make_nifti_image(np.arange(64).reshape(1, 8, 8), np.diag([1.5, 1.5, 1.5, 1]))
data, header = LoadNifti(as_closest_canonical=False)(test_image)
data, original_affine, new_affine = Spacing([0.8, 0.8, 0.8])(
data[None], header["affine"], interp_order="nearest"
)
data, _, new_affine = Orientation("ILP")(data, new_affine)
if os.path.exists(test_image):
os.remove(test_image)
write_nifti(data[0], test_image, new_affine, original_affine, interp_order="nearest", mode="border")
saved = nib.load(test_image)
saved_data = saved.get_fdata()
np.testing.assert_allclose(saved_data, np.arange(64).reshape(1, 8, 8), atol=1e-7)
if os.path.exists(test_image):
os.remove(test_image)
write_nifti(
data[0],
test_image,
new_affine,
original_affine,
interp_order="nearest",
mode="border",
output_shape=(1, 8, 8),
)
saved = nib.load(test_image)
saved_data = saved.get_fdata()
np.testing.assert_allclose(saved_data, np.arange(64).reshape(1, 8, 8), atol=1e-7)
if os.path.exists(test_image):
os.remove(test_image)
def binary_to_image(reference_image,
label,
dtype=np.uint16,
file_ext=".nii.gz",
use_itk=True):
start = time.time()
image_np, meta_dict = LoadImage()(reference_image)
label_np = np.fromfile(label, dtype=dtype)
logger.info(f"Image: {image_np.shape}")
logger.info(f"Label: {label_np.shape}")
label_np = label_np.reshape(image_np.shape, order="F")
logger.info(f"Label (reshape): {label_np.shape}")
output_file = tempfile.NamedTemporaryFile(suffix=file_ext).name
affine = meta_dict.get("affine")
if use_itk:
write_itk(label_np,
output_file,
affine=affine,
dtype=None,
compress=True)
else:
write_nifti(label_np, output_file, affine=affine)
logger.info(f"binary_to_image latency : {time.time() - start} (sec)")
return output_file
def test_write_4d(self):
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(6).reshape((1, 1, 3, 2))
write_nifti(img, image_name, affine=np.diag([1.4, 1]), target_affine=np.diag([1, 1.4, 1]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[[[0, 1], [2, 3], [4, 5]]]])
np.testing.assert_allclose(out.affine, np.diag([1, 1.4, 1, 1]))
image_name = os.path.join(out_dir, "test1.nii.gz")
img = np.arange(5).reshape((1, 1, 5, 1))
write_nifti(img, image_name, affine=np.diag([1, 1, 1, 3, 3]), target_affine=np.diag([1.4, 2.0, 2, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[[[0], [2], [4]]]])
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]))
def test_write_3d(self):
out_dir = tempfile.mkdtemp()
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(6).reshape((1, 2, 3))
write_nifti(img, image_name, affine=np.diag([1]), target_affine=np.diag([1.4]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[[0, 1, 2], [3, 4, 5]]])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
image_name = os.path.join(out_dir, "test1.nii.gz")
img = np.arange(5).reshape((1, 1, 5))
write_nifti(img, image_name, affine=np.diag([1, 1, 1, 3, 3]), target_affine=np.diag([1.4, 2.0, 2, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[[0, 2, 4]]])
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]))
shutil.rmtree(out_dir)
def test_write_2d(self):
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, 'test.nii.gz')
img = np.arange(6).reshape((2, 3))
write_nifti(img,
image_name,
affine=np.diag([1]),
target_affine=np.diag([1.4]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[0, 1, 2], [0, 0, 0]])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, 'test.nii.gz')
img = np.arange(5).reshape((1, 5))
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1, 3, 3]),
target_affine=np.diag([1.4, 2., 1, 3, 5]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [[0, 2, 4]])
np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 1, 1]))
def test_write_1d(self):
out_dir = tempfile.mkdtemp()
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1]),
target_affine=np.diag([1.4, 2.0, 1]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [0, 1, 3, 0])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
image_name = os.path.join(out_dir, "test1.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img, image_name, affine=[[1]], target_affine=[[1.4]])
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [0, 1, 3, 0])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
image_name = os.path.join(out_dir, "test2.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img,
image_name,
affine=np.diag([1.5, 1.5, 1.5]),
target_affine=np.diag([1.5, 1.5, 1.5]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), np.arange(5).reshape(-1))
np.testing.assert_allclose(out.affine, np.diag([1.5, 1, 1, 1]))
shutil.rmtree(out_dir)
def test_consistency(self):
np.set_printoptions(suppress=True, precision=3)
test_image = make_nifti_image(
np.arange(64).reshape(1, 8, 8), np.diag([1.5, 1.5, 1.5, 1]))
data, header = LoadImage(reader="NibabelReader",
as_closest_canonical=False)(test_image)
data, original_affine, new_affine = Spacing([0.8, 0.8,
0.8])(data[None],
header["affine"],
mode="nearest")
data, _, new_affine = Orientation("ILP")(data, new_affine)
if os.path.exists(test_image):
os.remove(test_image)
write_nifti(data[0],
test_image,
new_affine,
original_affine,
mode="nearest",
padding_mode="border")
saved = nib.load(test_image)
saved_data = saved.get_fdata()
np.testing.assert_allclose(saved_data,
np.arange(64).reshape(1, 8, 8),
atol=1e-7)
if os.path.exists(test_image):
os.remove(test_image)
write_nifti(
data[0],
test_image,
new_affine,
original_affine,
mode="nearest",
padding_mode="border",
output_spatial_shape=(1, 8, 8),
)
saved = nib.load(test_image)
saved_data = saved.get_fdata()
np.testing.assert_allclose(saved_data,
np.arange(64).reshape(1, 8, 8),
atol=1e-7)
if os.path.exists(test_image):
os.remove(test_image)
# test the case that only correct orientation but don't resample
write_nifti(data[0],
test_image,
new_affine,
original_affine,
resample=False)
saved = nib.load(test_image)
# compute expected affine
start_ornt = nib.orientations.io_orientation(new_affine)
target_ornt = nib.orientations.io_orientation(original_affine)
ornt_transform = nib.orientations.ornt_transform(
start_ornt, target_ornt)
data_shape = data[0].shape
expected_affine = new_affine @ nib.orientations.inv_ornt_aff(
ornt_transform, data_shape)
np.testing.assert_allclose(saved.affine, expected_affine)
if os.path.exists(test_image):
os.remove(test_image)
def test_write_1d(self):
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img,
image_name,
affine=np.diag([1, 1, 1]),
target_affine=np.diag([1.4, 2.0, 1]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [0, 1, 3, 0])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img, image_name, affine=[[1]], target_affine=[[1.4]])
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(), [0, 1, 3, 0])
np.testing.assert_allclose(out.affine, np.diag([1.4, 1, 1, 1]))
with tempfile.TemporaryDirectory() as out_dir:
image_name = os.path.join(out_dir, "test.nii.gz")
img = np.arange(5).reshape(-1)
write_nifti(img,
image_name,
affine=np.diag([1.5, 1.5, 1.5]),
target_affine=np.diag([1.5, 1.5, 1.5]))
out = nib.load(image_name)
np.testing.assert_allclose(out.get_fdata(),
np.arange(5).reshape(-1))
np.testing.assert_allclose(out.affine, np.diag([1.5, 1, 1, 1]))
s = [bb[0, 0] - 16, bb[0, 2] - 16, bb[0, 4] - 16]
e = [bb[0, 1] + 16, bb[0, 3] + 16, bb[0, 5] + 16]
# generate lesion patches based on the bounding boxes
data_out = AddChanneld(keys)(data)
data_out = SpatialCropd(keys, roi_start=s, roi_end=e)(data_out)
resize = Resized(keys, patch_size, mode=("trilinear", "nearest"))
data_out = resize(data_out)
patch_out = (
f"{folder}/patch/lesion_{s[0]}_{s[1]}_{s[2]}_{e[0]}_{e[1]}_{e[2]}_{name_id}"
)
label_out = (
f"{folder}/patch/labels_{s[0]}_{s[1]}_{s[2]}_{e[0]}_{e[1]}_{e[2]}_{name_id}"
)
write_nifti(data_out["image"][0], file_name=patch_out)
write_nifti(data_out["label"][0], file_name=label_out)
# generate random negative samples
rand_data_out = AddChanneld(keys)(data)
rand_data_out["inv_label"] = (rand_data_out["label"] == 0
) # non-lesion sampling map
rand_crop = RandWeightedCropd(keys,
"inv_label",
patch_size,
num_samples=3)
rand_data_out = rand_crop(rand_data_out)
for idx, d in enumerate(rand_data_out):
if np.sum(d["label"]) > 0:
continue
write_nifti(d["image"][0],
def predictionAndEvaluation(params, test_loader, device, model):
# TODO: affine matrices with different orientation -------------------------------------------------------------------------------
# TODO: 2D affine 'ra' or 'rs'
# TODO: respacing ---------------------------------
test_affine = getAffineMatrix(params['orientation'], params['spacing'], params['model']['dimensions'])
model_metrics = params['model']['metricfunction']
model_metrics['reduction'] = 'none'
eval_metric = eval(model_objects.concatFunctionArguments(model_metrics, prefix='monaiMetrics.'))
original_dice_results = list()
postprocess_dice_results = list()
output_results = list()
original_hd_results = list()
postprocess_hd_results = list()
original_mad_results = list()
postprocess_mad_results = list()
with torch.no_grad():
for i, test_data in enumerate(test_loader):
roi_size = params['model']['roi_size']
sw_batch_size = params['model']['sw_batch_size']
test_outputs = sliding_window_inference(test_data['IMAGE'].to(device), roi_size, sw_batch_size, model)
argmax = torch.argmax(test_outputs, dim=1, keepdim=True)
# Post-processing
if params['postprocessing']:
post_func = eval(model_objects.concatFunctionArguments(params['postprocessing_func'], prefix='monaiTrans.'))
largest = post_func(argmax)
else:
largest = argmax
# Write data out
output_file_name = os.path.split(test_loader.dataset.data[i]['IMAGE'])[1]
output_path = os.path.join(params['output_path'],output_file_name)
output_results.append(output_path)
test_data_nii = nib.load(test_loader.dataset.data[i]['IMAGE'])
target_affine = test_data_nii.affine
# Remove the translation component
# TODO: 2D-----------------------------------------------------------------------------------------
# TODO: xenos please take a look at it
target_affine[0:params['model']['dimensions'],params['model']['dimensions']] = 0
# TODO: xenos please take a look at it: resampling ------------------------------------------------------------------
monaiData.write_nifti(largest.detach().cpu()[0, 0,...].numpy(), output_path,
mode='nearest',
affine=test_affine,
target_affine=target_affine,
output_spatial_shape=test_data_nii.shape,
dtype=np.float32
)
if params['debug']:
print(test_data['IMAGE'].shape)
print(output_path)
# evaluation scores
if params['testSegm']:
test_labels = test_data['SEGM'].to(device)
value = eval_metric(y_pred=argmax, y=test_labels)
print('Dice: {:.5f}'.format(value.item()))
original_dice_results.append(value.item())
hd_value = monaiMetrics.compute_hausdorff_distance(argmax, test_labels, label_idx=1, percentile=95)
print('HD95: {:.5f}'.format(hd_value.item()))
original_hd_results.append(hd_value)
mad_value = monaiMetrics.compute_average_surface_distance(argmax, test_labels, label_idx=1)
print('MAD: {:.5f}'.format(mad_value.item()))
original_mad_results.append(mad_value)
if params['postprocessing']:
value = eval_metric(y_pred=largest, y=test_labels)
postprocess_dice_results.append(value.item())
print('Post-processed Dice: {:.5f}'.format(value.item()))
hd_value = monaiMetrics.compute_hausdorff_distance(largest, test_labels, label_idx=1, percentile=95)
print('Post-processed HD95: {:.5f}'.format(hd_value.item()))
postprocess_hd_results.append(hd_value)
mad_value = monaiMetrics.compute_average_surface_distance(largest, test_labels, label_idx=1)
print('Post-processed HD95: {:.5f}'.format(mad_value.item()))
postprocess_mad_results.append(mad_value)
if params['testSegm']:
eval_results = pd.DataFrame()
eval_results['IMAGE_DATA'] = [ i['IMAGE'] for i in test_loader.dataset.data]
eval_results['SEGM_DATA'] = [ i['SEGM'] for i in test_loader.dataset.data]
eval_results['SEGM_RESULTS'] = output_results
eval_results['DICE'] = original_dice_results
if params['postprocessing']:
eval_results['POST_DICE'] = postprocess_dice_results
eval_results['HD95'] = original_hd_results
if params['postprocessing']:
eval_results['POST_HD95'] = postprocess_hd_results
eval_results['MAD'] = original_mad_results
if params['postprocessing']:
eval_results['POST_MAD'] = postprocess_mad_results
eval_results.to_csv(os.path.join(params['output_path'],'evaluation_results.csv'), index=False)
