def sample_patches(lid, imgpath, lblpath):
image = sitk.ReadImage(imgpath)
image = resample(image, (1.0, 1.0, 1.0), interpolator = sitk.sitkLinear)
img_arr = sitk.GetArrayFromImage(image)
img_arr = snd.zoom(img_arr, zoom = (0.5, 0.5, 0.5), order = 1)
img_arr = np.float32(np.clip(img_arr, -100, 400))
img_arr = np.uint8(255*(img_arr + 100)/(500))
img_arr = np.pad(img_arr, ((100,100),(100,100),(100,100)), mode = 'constant')
label = sitk.ReadImage(lblpath)
label = resample(label, (1.0, 1.0, 1.0), interpolator = sitk.sitkNearestNeighbor)
lbl_arr = sitk.GetArrayFromImage(label)
lbl_arr[lbl_arr == 2] = 1
lbl_arr = np.uint8(snd.zoom(lbl_arr, zoom = (0.5, 0.5, 0.5), order = 0))
lbl_arr_cp = lbl_arr.copy() + 1
lbl_arr = np.pad(lbl_arr, ((100,100),(100,100),(100,100)), mode = 'constant')
lbl_arr_cp = np.pad(lbl_arr_cp, ((100,100),(100,100),(100,100)), mode = 'constant')
lbl_arr_cp -= 1
class1_locs = uniform_sample(lbl_arr_cp == 0, 50)
class2_locs = uniform_sample(lbl_arr_cp == 1, 50)
# print(' class 1, class 2 :', len(class1_locs), len(class2_locs))
locs = class1_locs[:5] + class2_locs[:45]
random.shuffle(locs)
patch_size, lbl_size = [116, 132, 132], [28, 44, 44]
liver_pixel_count = {}
for idx, l in enumerate(locs):
l = adjust_center_for_boundaries(l, patch_size, img_arr.shape)
img_patch = extract_patch(img_arr, l, patch_size)
lbl_patch = extract_patch(lbl_arr, l, lbl_size)
liver_pixel_count[idx] = np.sum(lbl_patch)
save_dir = './data/train'
inppname = 'img' + str(lid) + '_input'+str(idx)+'.npy'
tgtpname = 'img' + str(lid) + '_label'+str(idx)+'.npy'
np.save(os.path.join(save_dir, inppname), img_patch)
np.save(os.path.join(save_dir, tgtpname), lbl_patch)
def sample_patches(lid, imgpath, lblpath):
'''
Extract Patches
Parameters:
imgpath - path of the image
lblpath - path of the corressponding segmented image
lid -
'''
#Image
#Read the image and return an object (generator comprehension), Here 'image' is a sitk's object
image = sitk.ReadImage(imgpath)
image = resample(image, (1.0, 1.0, 1.0), interpolator=sitk.sitkLinear)
#get the pixel values from the image
img_arr = sitk.GetArrayFromImage(image)
#Zooms the image according to the zoom
img_arr = snd.zoom(img_arr, zoom=(0.5, 0.5, 0.5), order=1)
#pixel values are clipped b/w (-100, 400) and converted to float32
img_arr = np.float32(np.clip(img_arr, -100,
400)) #Why clipping b/w (-100,400)?
#Some kind of data preprocessing
img_arr = np.uint8(255 * (img_arr + 100) / (500))
#Pad the image
img_arr = np.pad(img_arr, ((100, 100), (100, 100), (100, 100)),
mode='constant')
#Label
label = sitk.ReadImage(lblpath)
label = resample(label, (1.0, 1.0, 1.0),
interpolator=sitk.sitkNearestNeighbor)
lbl_arr = sitk.GetArrayFromImage(label)
lbl_arr[lbl_arr == 2] = 1
lbl_arr = np.uint8(snd.zoom(lbl_arr, zoom=(0.5, 0.5, 0.5), order=0))
#Copies the content of 'lbl_arr' and adds '1' to it
lbl_arr_cp = lbl_arr.copy() + 1
lbl_arr = np.pad(lbl_arr, ((100, 100), (100, 100), (100, 100)),
mode='constant')
lbl_arr_cp = np.pad(lbl_arr_cp, ((100, 100), (100, 100), (100, 100)),
mode='constant')
lbl_arr_cp -= 1
#Getting the crops for liver class and non-liver class
class1_locs = uniform_sample(lbl_arr_cp == 0, 50)
class2_locs = uniform_sample(lbl_arr_cp == 1, 50)
# print(' class 1, class 2 :', len(class1_locs), len(class2_locs))
locs = class1_locs[:5] + class2_locs[:45]
random.shuffle(locs)
patch_size, lbl_size = [116, 132, 132], [28, 44, 44]
liver_pixel_count = {}
for idx, l in enumerate(locs):
l = adjust_center_for_boundaries(l, patch_size, img_arr.shape)
img_patch = extract_patch(img_arr, l, patch_size)
lbl_patch = extract_patch(lbl_arr, l, lbl_size)
liver_pixel_count[idx] = np.sum(lbl_patch)
save_dir = './data/train'
inppname = 'img' + str(lid) + '_input' + str(idx) + '.npy'
tgtpname = 'img' + str(lid) + '_label' + str(idx) + '.npy'
np.save(os.path.join(save_dir, inppname), img_patch)
np.save(os.path.join(save_dir, tgtpname), lbl_patch)
def inference(self, inputs):
if config.mode is "fcn":
fm = utils.conv_with_bn(inputs, out_channels=12, filter_size=[config.time_width, 13],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_1")
fm = utils.conv_with_bn(fm, out_channels=16, filter_size=[config.time_width, 11],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_2")
fm = utils.conv_with_bn(fm, out_channels=20, filter_size=[config.time_width, 9],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_3")
fm_skip = utils.conv_with_bn(fm, out_channels=24, filter_size=[config.time_width, 7],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_4")
fm = utils.conv_with_bn(fm_skip, out_channels=32, filter_size=[config.time_width, 7],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_5")
fm = utils.conv_with_bn(fm, out_channels=24, filter_size=[config.time_width, 7],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_6") + fm_skip
fm = utils.conv_with_bn(fm, out_channels=20, filter_size=[config.time_width, 9],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_7")
fm = utils.conv_with_bn(fm, out_channels=16, filter_size=[config.time_width, 11],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_8")
fm = utils.conv_with_bn(fm, out_channels=12, filter_size=[config.time_width, 13],
stride=1, act='relu', is_training=self._is_training,
padding="SAME", name="conv_9")
fm = utils.conv_with_bn(fm, out_channels=1, filter_size=[config.time_width, config.freq_size],
stride=1, act='linear', is_training=self._is_training,
padding="SAME", name="conv_10") # (batch_size, 1, config.freq_size, 1)
# fm = utils.conv_with_bn(fm, out_channels=1, filter_size=[config.time_width, 1],
# stride=1, act='linear', is_training=self._is_training,
# padding="VALID", name="conv_last")
fm = tf.squeeze(fm, [1, 3])
return fm
elif config.mode is "fnn":
keep_prob = self.keep_prob
# inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
# inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
#
# h1 = tf.nn.relu(utils.batch_norm_affine_transform(inputs, 2048, name='hidden_1',
# is_training=self._is_training))
# # h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
#
# h2 = tf.nn.relu(utils.batch_norm_affine_transform(h1, 2048, name='hidden_2',
# is_training=self._is_training))
# # h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
#
# # h3 = tf.nn.relu(utils.batch_norm_affine_transform(h2, 2048, name='hidden_3',
# # is_training=self._is_training))
# # h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
#
# fm = utils.affine_transform(h2, config.freq_size, name='logits')
inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
h1 = tf.nn.selu(utils.affine_transform(inputs, 2048, name='hidden_1'))
h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
h2 = tf.nn.selu(utils.affine_transform(h1, 2048, name='hidden_2'))
h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
h3 = tf.nn.selu(utils.affine_transform(h2, 2048, name='hidden_3'))
h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
fm = utils.affine_transform(h3, config.freq_size, name='logits')
return fm
elif config.mode is "irm":
keep_prob = self.keep_prob
# inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
# inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
#
# h1 = tf.nn.relu(utils.batch_norm_affine_transform(inputs, 2048, name='hidden_1',
# is_training=self._is_training))
# # h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
#
# h2 = tf.nn.relu(utils.batch_norm_affine_transform(h1, 2048, name='hidden_2',
# is_training=self._is_training))
# # h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
#
# # h3 = tf.nn.relu(utils.batch_norm_affine_transform(h2, 2048, name='hidden_3',
# # is_training=self._is_training))
# # h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
#
# fm = utils.affine_transform(h2, config.freq_size, name='logits')
inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
h1 = tf.nn.selu(utils.affine_transform(inputs, 2048, name='hidden_1'))
h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
h2 = tf.nn.selu(utils.affine_transform(h1, 2048, name='hidden_2'))
h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
h3 = tf.nn.selu(utils.affine_transform(h2, 2048, name='hidden_3'))
h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
fm = utils.affine_transform(h3, config.freq_size, name='logits')
return fm
elif config.mode is "sfnn":
keep_prob = self.keep_prob
skip_inputs = tf.squeeze(inputs[:, int(config.time_width/2), :])
inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
h1 = tf.nn.selu(utils.affine_transform(inputs, 2048, name='hidden_1'))
h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
h2 = tf.nn.selu(utils.affine_transform(h1, 2048, name='hidden_2'))
h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
h3 = tf.nn.selu(utils.affine_transform(h2, 2048, name='hidden_3'))
h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
fm = utils.affine_transform(h3, config.freq_size, name='logits')
fm = fm + skip_inputs
return fm
elif config.mode is "lstm":
keep_prob = self.keep_prob
# inputs = tf.squeeze(inputs)[:, int(config.time_width/2), :]
# inputs = tf.reshape(inputs, (-1, config.time_width, config.freq_size)) # time_width == num_steps
# inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
num_units = [1024, 1024]
cells = [tf.nn.rnn_cell.LSTMCell(num_units=n, state_is_tuple=True) for n in num_units]
cell = tf.nn.rnn_cell.MultiRNNCell(cells=cells, state_is_tuple=True)
cell = tf.contrib.rnn.OutputProjectionWrapper(cell, output_size=config.freq_size)
outputs, _state = tf.nn.dynamic_rnn(cell, inputs, time_major=False, dtype=tf.float32)
fm = tf.reshape(outputs,[-1, config.freq_size])
return fm
elif config.mode is "tsn":
conv_inputs = tf.squeeze(tf.transpose(inputs, [0, 2, 1, 3]), axis=3)
keep_prob = self.keep_prob
# inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width*config.freq_size)))
# inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
#
# h1 = tf.nn.relu(utils.batch_norm_affine_transform(inputs, 2048, name='hidden_1',
# is_training=self._is_training))
# # h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
#
# h2 = tf.nn.relu(utils.batch_norm_affine_transform(h1, 2048, name='hidden_2',
# is_training=self._is_training))
# # h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
#
# # h3 = tf.nn.relu(utils.batch_norm_affine_transform(h2, 2048, name='hidden_3',
# # is_training=self._is_training))
# # h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
#
# fm = utils.affine_transform(h2, config.freq_size, name='logits')
skip_inputs = tf.squeeze(inputs)[:, int(config.time_width / 2), :]
skip_inputs = tf.squeeze(inputs, axis=3)
inputs = tf.reshape(tf.squeeze(inputs, [3]), (-1, int(config.time_width * config.freq_size)))
inputs = tf.nn.dropout(inputs, keep_prob=keep_prob)
h1 = tf.nn.selu(utils.affine_transform(inputs, 1024, name='hidden_1'))
h1 = tf.nn.dropout(h1, keep_prob=keep_prob)
h2 = tf.nn.selu(utils.affine_transform(h1, 1024, name='hidden_2'))
h2 = tf.nn.dropout(h2, keep_prob=keep_prob)
h3 = tf.nn.selu(utils.affine_transform(h2, 1024, name='hidden_3'))
h3 = tf.nn.dropout(h3, keep_prob=keep_prob)
fm = utils.affine_transform(h3, int(config.freq_size * config.time_width), name='logits')
fm = tf.reshape(fm, (-1, config.time_width, config.freq_size))
pad = tf.zeros((1, config.freq_size * int(config.time_width / 2), config.time_width, 1))
conv_fm = tf.reshape(tf.transpose(tf.expand_dims(fm, axis=3), [0, 2, 1, 3]),
(1, -1, config.time_width, 1))
conv_fm = tf.concat([pad, conv_fm, pad], axis=1)
conv_fm = utils.extract_patch(tf.squeeze(conv_fm),
patch_size=(config.freq_size * config.time_width, config.time_width))
conv_fm = tf.stack(tf.split(conv_fm, num_or_size_splits=config.time_width, axis=1), axis=3)
conv_fm = tf.reshape(conv_fm, (-1, config.freq_size, config.time_width * config.time_width))
# att_inputs = tf.reshape(conv_fm, (-1, config.freq_size*config.time_width*config.time_width))
# h4 = tf.nn.selu(utils.affine_transform(att_inputs, 1024, name='hidden_4'))
# h4 = tf.nn.dropout(h4, keep_prob=keep_prob)
# h5 = tf.nn.selu(utils.affine_transform(h4, config.time_width*config.time_width, name='hidden_5'))
# att_outputs = tf.expand_dims(tf.nn.softmax(h5), axis=1)
conv_fm = tf.concat([conv_fm, conv_inputs], axis=2)
conv_fm = tf.expand_dims(conv_fm, axis=2)
conv_1 = utils.conv_with_bn_2(conv_fm, 256, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_1')
conv_2 = utils.conv_with_bn_2(conv_1, 128, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_2')
conv_3 = utils.conv_with_bn_2(conv_2, 64, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_3')
conv_4 = utils.conv_with_bn_2(conv_3, 32, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_4')
conv_5 = utils.conv_with_bn_2(conv_4, 32, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_5')
conv_6 = utils.conv_with_bn_2(conv_5, 32, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_6')
conv_7 = utils.conv_with_bn_2(conv_6, 32, filter_size=[5, 1], stride=1, act='relu', scale=True,
is_training=self._is_training, padding="SAME", name='conv_7')
conv_8 = utils.conv_with_bn_2(conv_7, 1, filter_size=[5, 1], stride=1, act='relu', scale=False,
is_training=self._is_training, padding="SAME", name='conv_8')
conv_9 = tf.squeeze(tf.squeeze(conv_8, axis=2), axis=2)
return fm, conv_9
def main():
parser = argparse.ArgumentParser(
description="Extract the 3D patches containing the nodules and store them in TFRecord files.",
)
parser.add_argument(
"data_dir",
help="Directory containing all the DCM files downloaded from https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge",
)
parser.add_argument(
"test_xls_file",
help="Test Excel file obtained from https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge",
)
# parser.add_argument(
# "calibration_xls_file",
# help="Calibration Excel file obtained from https://wiki.cancerimagingarchive.net/display/Public/SPIE-AAPM+Lung+CT+Challenge",
# )
args = parser.parse_args()
data_dir = Path(args.data_dir)
df = read_xls(args.test_xls_file)
# test_df = read_xls(args.test_xls_file)
# calibration_df = read_xls(args.calibration_xls_file)
# df = pd.concat([test_df, calibration_df])
assert len(df.index) == 73, "The input excels have not the expected size."
with tf.io.TFRecordWriter(
SPIE_SMALL_NEG_TFRECORD
) as small_neg_writer, tf.io.TFRecordWriter(
SPIE_SMALL_POS_TFRECORD
) as small_pos_writer, tf.io.TFRecordWriter(
SPIE_BIG_NEG_TFRECORD
) as big_neg_writer, tf.io.TFRecordWriter(
SPIE_BIG_POS_TFRECORD
) as big_pos_writer:
for row in tqdm(df.itertuples(), total=len(df.index)):
dcm_dir_glob = Path(data_dir).glob(f"{row.scan_number}/*/*/")
dcm_dir = list(dcm_dir_glob)[0]
scan = read_dcm(dcm_dir, reverse_z=False)
big_writer = big_pos_writer if row.label else big_neg_writer
big_patch = extract_patch(
scan,
(row.zloc, row.yloc, row.xloc),
BIG_PATCH_SHAPE[:-1],
)
big_patch = pad_to_shape(big_patch, BIG_PATCH_SHAPE[:-1])
big_patch = np.expand_dims(big_patch, axis=-1)
if not big_patch.any():
print(
f"WARNING ({row.scan_number=}): "
"Patch contains only zeros. Skipping this patch ..."
)
continue
assert (
big_patch.shape == BIG_PATCH_SHAPE
), f"Wrong shape for scan {row.scan_number}."
big_patch = big_patch.astype(np.float32)
big_example = volume_to_example(big_patch)
big_writer.write(big_example.SerializeToString())
small_writer = small_pos_writer if row.label else small_neg_writer
small_patch = extract_patch(
scan,
(row.zloc, row.yloc, row.xloc),
SMALL_PATCH_SHAPE[:-1],
)
small_patch = pad_to_shape(small_patch, SMALL_PATCH_SHAPE[:-1])
small_patch = np.expand_dims(small_patch, axis=-1)
assert (
small_patch.shape == SMALL_PATCH_SHAPE
), f"Wrong shape for scan {row.scan_number}."
small_patch = small_patch.astype(np.float32)
small_example = volume_to_example(small_patch)
small_writer.write(small_example.SerializeToString())
def main():
parser = argparse.ArgumentParser(
description=
"Extract the 3D patches containing the nodules and store them in TFRecord files.",
)
parser.add_argument(
"data_dir",
help=
"Directory containing all the DCM files downloaded from https://wiki.cancerimagingarchive.net/display/Public/LIDC-IDRI",
)
parser.add_argument(
"csv_file",
help="CSV file obtained from http://www.via.cornell.edu/lidc",
)
args = parser.parse_args()
data_dir = Path(args.data_dir)
nodules_df = read_lidc_size_report(args.csv_file)
assert (len(nodules_df.index) == 1387
), f"The input CSV {args.csv_file} has not the expected size."
with tf.io.TFRecordWriter(
LIDC_SMALL_NEG_TFRECORD) as small_neg_writer, tf.io.TFRecordWriter(
LIDC_SMALL_POS_TFRECORD
) as small_pos_writer, tf.io.TFRecordWriter(
LIDC_BIG_NEG_TFRECORD) as big_neg_writer, tf.io.TFRecordWriter(
LIDC_BIG_POS_TFRECORD
) as big_pos_writer, tf.io.TFRecordWriter(
LIDC_SMALL_UNLABELED_TFRECORD
) as small_unlabeled_writer, tf.io.TFRecordWriter(
LIDC_BIG_UNLABELED_TFRECORD) as big_unlabeled_writer:
for row in tqdm(nodules_df.itertuples(), total=len(nodules_df.index)):
case = row.case
scan_id = row.scan
dcm_dir_glob = list(
data_dir.glob(f"LIDC-IDRI-{case}/*/{scan_id}.*/"))
if len(dcm_dir_glob) == 0:
print(f"WARNING ({scan_id=} {case=}): "
"Scan not found. Skipping this scan ...")
continue
if len(dcm_dir_glob) > 1:
print(
f"WARNING ({scan_id=} {case=}): "
"Found multiple scans with same ids. Skipping this scan ..."
)
continue
dcm_dir = dcm_dir_glob[0]
scan = read_dcm(dcm_dir, reverse_z=True)
xml_files = list(dcm_dir.glob("*.xml"))
if len(xml_files) == 0:
print(f"WARNING ({scan_id=} {case=}): "
"Can't find a XML file. Skipping this scan ...")
continue
elif len(xml_files) > 1:
print(f"WARNING ({scan_id=} {case=}): "
"Found multiple XML files. Skipping this scan ...")
continue
xml_file = xml_files[0]
nodule_ids = row.ids
malignancies = get_malignancies(xml_file, nodule_ids)
median_malignancy = median(malignancies)
if median_malignancy < 3:
big_writer = big_neg_writer
small_writer = small_neg_writer
elif median_malignancy > 3:
big_writer = big_pos_writer
small_writer = small_pos_writer
else:
# if the malignancies median is 3 then write the patch
# as unlabeled
big_writer = big_unlabeled_writer
small_writer = small_unlabeled_writer
big_patch = extract_patch(
scan,
(row.zloc, row.yloc, row.xloc),
BIG_PATCH_SHAPE[:-1],
)
big_patch = pad_to_shape(big_patch, BIG_PATCH_SHAPE[:-1])
big_patch = np.expand_dims(big_patch, axis=-1)
if not big_patch.any():
print(f"WARNING ({scan_id=} {case=}): "
"Patch contains only zeros. Skipping this patch ...")
continue
assert (big_patch.shape == BIG_PATCH_SHAPE
), f"Wrong shape for scan {scan_id} in case {case}."
big_patch = big_patch.astype(np.float32)
big_example = volume_to_example(big_patch)
big_writer.write(big_example.SerializeToString())
small_patch = extract_patch(
scan,
(row.zloc, row.yloc, row.xloc),
SMALL_PATCH_SHAPE[:-1],
)
small_patch = pad_to_shape(small_patch, SMALL_PATCH_SHAPE[:-1])
small_patch = np.expand_dims(small_patch, axis=-1)
assert (small_patch.shape == SMALL_PATCH_SHAPE
), f"Wrong shape for scan {scan_id} in case {case}."
small_patch = small_patch.astype(np.float32)
small_example = volume_to_example(small_patch)
small_writer.write(small_example.SerializeToString())