def __init__(self, file_name, start_iter=0,
stop_iter=None):
self.file_name = file_name
if '.nd2' in file_name:
self.ext = 'nd2'
self.reader = ND2Reader(file_name)
elif ('.tif' in file_name) or ('.tiff' in file_name):
self.ext = 'tif'
self.reader = tifffile.TiffFile(file_name)
elif ('.czi' in file_name):
self.ext = 'czi'
self.reader = CziFile(file_name)
else:
raise RuntimeError("Image format %s " \
"not recognized" % \
os.path.splitext(file_name)[1])
# Record the shape of the movie
self.n_frames, self.height, self.width = \
self.get_shape()
# Set the defaults for iterating over this
# object
self.start_iter = start_iter
self.stop_iter = stop_iter
if self.stop_iter is None:
self.stop_iter = self.n_frames
def open(self, location, **kwargs):
with warnings.catch_warnings():
warnings.simplefilter('ignore')
self.czi = CziFile(location.path)
self.frames = {
_get_subblock_identifier(subblock): subblock
for subblock in self.czi.filtered_subblock_directory
}
self.size = _dim_shape_to_dict(self.czi.axes, self.czi.shape)
self.metadata = self.czi.metadata(raw=False)
scaling = self.metadata['ImageDocument']['Metadata']['Scaling'][
'Items']['Distance']
# /ImageDocument
calibration_x = float(
next(iter([scale for scale in scaling if scale['Id'] == 'X'
]))['Value']) * 1E6
calibration_y = float(
next(iter([scale for scale in scaling if scale['Id'] == 'Y'
]))['Value']) * 1E6
assert calibration_x == calibration_y
self.calibration = calibration_x
timestamps = None
for entry in self.czi.attachment_directory:
if entry.name == 'TimeStamps':
timestamps = entry.data_segment().data()
break
self.timestamps = timestamps
positions = []
for scene in sorted(
self.metadata['ImageDocument']['Metadata']['Information']
['Image']['Dimensions']['S']['Scenes']['Scene'],
key=lambda scene_: int(scene_['Index'])):
x, y = scene['CenterPosition'].split(',')
positions.append((float(x), float(y)))
self.positions = positions
self.set_dimensions_and_sizes([
Dimensions.Time, Dimensions.PositionXY, Dimensions.PositionZ,
Dimensions.Channel
], [
self.size.get('T', 1),
self.size.get('S', 1), 1,
self.size.get('C', 1)
])
def getMetadata(self):
'''
Open czi xml header
convert to dictionary
extract needed information and convert to correct units
return new dictionary
'''
with CziFile(self.directory) as c:
cziMetadata = c.metadata()
cziMetadata = parse(cziMetadata)
#X, Y, & Z pixel size converted from meters to microns
x_pixel_size = (10**6)*float(cziMetadata["ImageDocument"]["Metadata"]["Scaling"]["Items"]["Distance"][0]["Value"])
y_pixel_size = (10**6)*float(cziMetadata["ImageDocument"]["Metadata"]["Scaling"]["Items"]["Distance"][1]["Value"])
z_pixel_size = (10**6)*float(cziMetadata["ImageDocument"]["Metadata"]["Scaling"]["Items"]["Distance"][2]["Value"])
voxel_size = (x_pixel_size * y_pixel_size * z_pixel_size)
self.metadata = {'x_pixel_size' : x_pixel_size,
'y_pixel_size' : y_pixel_size,
'z_pixel_size' : z_pixel_size,
'voxel_size' : voxel_size}
return self.metadata
def load_zeiss(path):
_, img_name = os.path.split(path)
with CziFile(path) as czi:
xmltxt = czi.metadata()
meta = xml.etree.ElementTree.fromstring(xmltxt)
# next line is somewhat cryptic, but just extracts um/pix (calibration) of X and Y into res
res = [float(i[0].text) for i in meta.findall('.//Scaling/Items/*') if
i.attrib['Id'] == 'X' or i.attrib['Id'] == 'Y']
assert np.isclose(res[0], res[1]), "Pixels are not square."
# get first calibration value and convert it from meters to um
res = res[0] * 1e6
ts_ix = [k for k, a1 in enumerate(czi.attachment_directory) if a1.filename[:10] == 'TimeStamps'][0]
timestamps = list(czi.attachments())[ts_ix].data()
dt = np.median(np.diff(timestamps))
ax_dct = {n: k for k, n in enumerate(czi.axes)}
n_frames = czi.shape[ax_dct['T']]
n_channels = czi.shape[ax_dct['C']]
n_X = czi.shape[ax_dct['X']]
n_Y = czi.shape[ax_dct['Y']]
images = list()
for sb in czi.subblock_directory:
images.append(sb.data_segment().data().reshape((n_X, n_Y)))
return MetadataImage(image=np.array(images), pix_per_um=1. / res, um_per_pix=res, time_interval=dt,
frames=n_frames, channels=n_channels,
width=n_X, height=n_Y, series=None)
def openFile(img_f):
from czifile import CziFile
with CziFile(img_f) as czi:
image_arrays = czi.asarray()
print 'Shape:', image_arrays.shape
# Returns several dimensions:
# ?, colors, ?, Z, X, Y, ?
return image_arrays
def read_czifile(czi_file, squeeze=True):
"""
Reads czifile
"""
from czifile import CziFile
with CziFile(czi_file) as czi:
image_arrays = czi.asarray()
if squeeze:
image_arrays = np.squeeze(image_arrays)
return image_arrays
def get_czi_metadata(p_czi, p_out=None, stdout=False):
"""
Convenience function to strip the metadata out of a CZI image and return. \
Requires CZI path. If `p_out` is provided, log the output to file. If \
`stdout` is True, print the metadata to standard output.
"""
with CziFile(p_czi) as czi:
meta = czi.metadata()
if p_out:
with open(p_out, 'w') as outfile:
outfile.writelines(meta)
if stdout:
print(meta)
return meta
def read_czi_image(czi_file, channel_names=None):
from czifile import CziFile
czi_img = CziFile(czi_file)
czi_channels = np.transpose(czi_img.asarray()[0, :, 0, :, :, :, 0],
(0, 2, 3, 1))
voxelsize = {}
for s in czi_img.segments():
if s.SID == "ZISRAWMETADATA":
metadata = s.data().split('\n')
for i, row in enumerate(metadata):
if "Distance Id" in row:
s = metadata[i + 1]
voxelsize[row.split('"')[1]] = np.around(
float(s[s.find('>') + 1:s.find('>') +
s[s.find('>'):].find('<')]) * 1e6,
decimals=3)
voxelsize = tuple([voxelsize[dim] for dim in ['X', 'Y', 'Z']])
n_channels = czi_channels.shape[0]
print czi_file.split('/')[-1], " : ", n_channels, " Channels ", voxelsize
if n_channels > 1:
if channel_names is None:
channel_names = ["CH" + str(i) for i in range(n_channels)]
img = {}
for i_channel, channel_name in enumerate(channel_names):
img[channel_name] = SpatialImage(czi_channels[i_channel],
voxelsize=voxelsize)
else:
img = SpatialImage(czi_channels[0], voxelsize=voxelsize)
return img
def test_czifile():
"""Test JpegXR compressed CZI file."""
from czifile import CziFile
fname = datafiles('jpegxr.czi')
if not os.path.exists(fname):
pytest.skip('large file not included with source distribution')
with CziFile(fname) as czi:
assert czi.shape == (1, 1, 15, 404, 356, 1)
assert czi.axes == 'BCZYX0'
# verify data
data = czi.asarray()
assert data.flags['C_CONTIGUOUS']
assert data.shape == (1, 1, 15, 404, 356, 1)
assert data.dtype == 'uint16'
assert data[0, 0, 14, 256, 146, 0] == 38086
def __init__(self,fpath,summary=True):
'''
Read in file using czifile
Parameters
----------
fpath : str
Complete or relative file path to czi file
'''
with CziFile(fpath) as czi:
self.raw_im = czi.asarray()
if summary:
self.print_summary()
self.squeeze_data()
def get_im_info(image_filepath):
"""
Use CziFile and tifffile to get image dimension and other information.
Parameters
----------
image_filepath: str
filepath to the image file
Returns
-------
im_dims: np.ndarray
image dimensions in np.ndarray
im_dtype: np.dtype
data type of the image
reader:str
whether to use "czifile" or "tifffile" to read the image
"""
if Path(image_filepath).suffix.lower() == ".czi":
czi = CziFile(image_filepath)
ch_dim_idx = czi.axes.index('C')
y_dim_idx = czi.axes.index('Y')
x_dim_idx = czi.axes.index('X')
im_dims = np.array(czi.shape)[[ch_dim_idx, y_dim_idx, x_dim_idx]]
im_dtype = czi.dtype
reader = "czi"
elif Path(image_filepath).suffix.lower() in TIFFFILE_EXTS:
largest_series = tf_get_largest_series(image_filepath)
zarr_im = zarr.open(
imread(image_filepath, aszarr=True, series=largest_series))
zarr_im = zarr_get_base_pyr_layer(zarr_im)
im_dims = np.squeeze(zarr_im.shape)
if len(im_dims) == 2:
im_dims = np.concatenate([[1], im_dims])
im_dtype = zarr_im.dtype
reader = "tifffile"
else:
im_dims, im_dtype = get_sitk_image_info(image_filepath)
reader = "sitk"
return im_dims, im_dtype, reader
def load_zeiss(path):
_, img_name = os.path.split(path)
with CziFile(path) as czi:
xmltxt = czi.metadata()
meta = xml.etree.ElementTree.fromstring(xmltxt)
# next line is somewhat cryptic, but just extracts um/pix (calibration) of X and Y into res
res = [
float(i[0].text) for i in meta.findall('.//Scaling/Items/*')
if i.attrib['Id'] == 'X' or i.attrib['Id'] == 'Y'
]
assert np.isclose(res[0], res[1]), "Pixels are not square."
# get first calibration value and convert it from meters to um
res = res[0] * 1e6
ts_ix = [
k for k, a1 in enumerate(czi.attachment_directory)
if a1.filename[:10] == 'TimeStamps'
][0]
timestamps = list(czi.attachments())[ts_ix].data()
dt = np.median(np.diff(timestamps))
ax_dct = {n: k for k, n in enumerate(czi.axes)}
n_frames = czi.shape[ax_dct['T']]
n_channels = czi.shape[ax_dct['C']]
n_zstacks = czi.shape[ax_dct['Z']]
n_x = czi.shape[ax_dct['X']]
n_y = czi.shape[ax_dct['Y']]
images = list()
for sb in czi.subblock_directory:
images.append(sb.data_segment().data().reshape((n_x, n_y)))
logger.info(
f"Loaded {czi._fh.name}. WxH({n_x},{n_y}), channels: {n_channels}, frames: {n_frames}, stacks: {n_zstacks}"
)
return np.array(
images), 1 / res, dt, n_frames, n_channels # , n_zstacks
class CZI:
def __init__(self, path):
self.filename = path
self.czi = CziFile(path)
def getPixelArray(self):
im = self.czi.asarray()
print('Image Width: {}'.format(im.shape[-1]))
print('Image Height: {}'.format(im.shape[-2]))
return im[0, 0, 0, ]
def cvtStandardImgFormat(self, savePath, fmt, compression=False):
im_arr = self.getPixelArray()
if fmt.endswith('jpg') and im_arr[0] > 65535 and im_arr[1] > 65535:
print('Too Large size for JPEG-2000 format...')
return
elif fmt.endswith('png') and im_arr[0] > 10000 and im_arr[1] > 10000:
print('Too Large size for PNG format...')
return
fname = os.path.basename(self.filename)
sname = os.path.splitext(fname)[-1]
im = Image.new_from_array(im_arr)
if fmt.endswith('tiff'):
if compression:
im.write_to_file(savePath + '/' + sname + '.' + fmt,
compression='lzw')
else:
im.write_to_file(savePath + '/' + sname + '.' + fmt,
compression='lzw')
else:
im.write_to_file(savePath + '/' + sname + '.' + fmt)
def read_z_slices_per_file(self, file_name: Path) -> int:
with CziFile(file_name) as czi_file:
return CziImageReader.find_count_of_axis(czi_file, "Z")
def read_sample_count(self, file_name: Path) -> int:
with CziFile(file_name) as czi_file:
return CziImageReader.find_count_of_axis(czi_file, "0")
def read_dimensions(self, file_name: Path) -> Tuple[int, int]:
with CziFile(file_name) as czi_file:
return (
CziImageReader.find_count_of_axis(czi_file, "X"),
CziImageReader.find_count_of_axis(czi_file, "Y"),
)
def read_array(
self,
file_name: Path,
dtype: Union[type, np.dtype],
z_slice: int,
channel_index: Optional[int],
sample_index: Optional[int],
) -> np.ndarray:
with CziFile(file_name) as czi_file:
# Read metadata
tile_shape = czi_file.filtered_subblock_directory[ # pylint: disable=unsubscriptable-object
0
].shape
dataset_shape: Tuple[int, ...] = tuple(czi_file.shape)
count = 0
used_axes = {"Z", "C", "Y", "X", "0"}
axes: Dict[str, Tuple[bool, int, int]] = {}
for i, axis in enumerate(czi_file.axes):
axes[axis] = (True, i, i - count)
if axis not in used_axes:
count += 1
for axis in used_axes:
if axis not in axes:
axes[axis] = (False, 0, 0)
assert tile_shape is not None, "Cannot read tile shape format."
if sample_index is not None:
num_output_channel = 1
elif channel_index is not None:
num_output_channel = dataset_shape[axes["0"][1]]
else:
channel_count = dataset_shape[axes["C"][1]] if axes["C"][0] else 1
num_output_channel = channel_count * dataset_shape[axes["0"][1]]
output_shape = (
dataset_shape[axes["X"][1]],
dataset_shape[axes["Y"][1]],
num_output_channel,
)
output = np.empty(output_shape, dtype)
z_file_start = (
czi_file.start[axes["Z"][1]] # pylint: disable=unsubscriptable-object
if axes["Z"][0]
else 0
)
c_file_start = (
czi_file.start[axes["C"][1]] # pylint: disable=unsubscriptable-object
if axes["C"][0]
else 0
)
output_channel_offset = 0
# Read the data
for tile_channel_index, (
tile_z_index,
data_z_index,
data_channel_index,
data_sample_index,
) in CziImageReader._select_correct_tiles_for_czi(
channel_index, sample_index, z_slice, axes, tile_shape, dataset_shape
).items():
# since the czi tiles are not sorted, we search linearly through them and check if the tile matches with the wanted coordinates
# however, some axes might not exist, so the _matches_if_exist helper returns true if the axis does not exist
for (
entry
) in (
czi_file.filtered_subblock_directory # pylint: disable=not-an-iterable
):
if CziImageReader._matches_if_exist(
"Z",
entry.start[axes["Z"][1]] - z_file_start,
tile_z_index,
axes,
) and CziImageReader._matches_if_exist(
"C",
(entry.start[axes["C"][1]] - c_file_start)
// tile_shape[axes["C"][1]],
tile_channel_index,
axes,
):
data = entry.data_segment().data()
data = to_target_datatype(data, dtype)
# left axis are in order [(Z)(C)0XY]
for axis in czi_file.axes: # pylint: disable=not-an-iterable
if axis not in used_axes:
data = data.take([0], axes[axis][2])
if data.ndim == 5:
data = data.transpose(
(
axes["Z"][2],
axes["X"][2],
axes["Y"][2],
axes["0"][2],
axes["C"][2],
)
)
elif data.ndim == 4:
if axes["Z"][0]:
data = data.transpose(
(
axes["Z"][2],
axes["X"][2],
axes["Y"][2],
axes["0"][2],
)
)
else:
data = data.transpose(
(
axes["X"][2],
axes["Y"][2],
axes["0"][2],
axes["C"][2],
)
)
elif data.ndim == 3:
data = data.transpose(
(
axes["X"][2],
axes["Y"][2],
axes["0"][2],
)
)
else:
raise Exception
if data_z_index is not None:
data = data.take([data_z_index], 0)
# if z axis exist, data_z_index is set, so the access index for this is always 0
if data_channel_index is not None and axes["C"][0]:
data = data.take([data_channel_index], -1)
# if the sample index is set, the channel index is set, too, so access index is always 0
if data_sample_index is not None:
data = data.take([data_sample_index], -1)
# special case with C and 0 and no selected index set
if data.ndim == 4:
for c in range(tile_shape[axes["C"][1]]):
next_output_channel_offset = (
output_channel_offset + data.shape[2]
)
output[
:,
:,
output_channel_offset:next_output_channel_offset,
] = data[:, :, :, c]
output_channel_offset = next_output_channel_offset
elif data.ndim == 2:
output[:, :, output_channel_offset] = data
output_channel_offset += 1
else:
output_channel_increment = (
1 if data_sample_index is not None else data.shape[-1]
)
next_output_channel_offset = (
output_channel_offset + output_channel_increment
)
output[
:, :, output_channel_offset:next_output_channel_offset
] = data
output_channel_offset = next_output_channel_offset
# CZI stores pixels as BGR instead of RGB, so swap axes to ensure right color output
if (
num_output_channel == 3
and dtype == np.uint8
and czi_file.filtered_subblock_directory[ # pylint: disable=unsubscriptable-object
0
].pixel_type
== "Bgr24"
):
output[:, :, 2], output[:, :, 0] = output[:, :, 0], output[:, :, 2]
output = output.reshape(output.shape + (1,))
return output
class ImageFileReader(object):
def __init__(self, file_name, start_iter=0,
stop_iter=None):
self.file_name = file_name
if '.nd2' in file_name:
self.ext = 'nd2'
self.reader = ND2Reader(file_name)
elif ('.tif' in file_name) or ('.tiff' in file_name):
self.ext = 'tif'
self.reader = tifffile.TiffFile(file_name)
elif ('.czi' in file_name):
self.ext = 'czi'
self.reader = CziFile(file_name)
else:
raise RuntimeError("Image format %s " \
"not recognized" % \
os.path.splitext(file_name)[1])
# Record the shape of the movie
self.n_frames, self.height, self.width = \
self.get_shape()
# Set the defaults for iterating over this
# object
self.start_iter = start_iter
self.stop_iter = stop_iter
if self.stop_iter is None:
self.stop_iter = self.n_frames
def __iter__(self):
self.c_idx = self.start_iter
return self
def __next__(self):
if self.c_idx < self.stop_iter:
self.c_idx += 1
return self.get_frame(self.c_idx-1)
else:
raise StopIteration
@property
def shape(self):
return self.get_shape()
@property
def dtype(self):
return self.get_frame(0).dtype
def _frame_valid(self, frame_idx):
return (frame_idx < self.n_frames)
def _frame_range_valid(self, frame_0, frame_1):
return (frame_0 <= self.n_frames) and \
(frame_1 <= self.n_frames) and \
(frame_0 < frame_1)
def _process_frame_range(self, start_frame, stop_frame):
if start_frame is None:
start_frame = 0
if stop_frame is None:
stop_frame = self.n_frames
assert self._frame_range_valid(start_frame, stop_frame)
return start_frame, stop_frame
def _subregion_valid(self, y0, y1, x0, x1):
return (y0<=self.height) and (y1<=self.height) \
and (x0<=self.width) and (x1<=self.width) \
and (y0 < y1) and (x0 < x1)
def _process_subregion(self, y0, y1, x0, x1):
if y0 is None:
y0 = 0
if x0 is None:
x0 = 0
if y1 is None:
y1 = self.height
if x1 is None:
x1 = self.width
assert self._subregion_valid(y0, y1, x0, x1)
return y0, y1, x0, x1
def close(self):
self.reader.close()
def get_shape(self):
"""
returns
-------
(int n_frames,
int height,
int width)
"""
if self.ext == 'nd2':
H = self.reader.metadata['height']
W = self.reader.metadata['width']
T = self.reader.metadata['total_images_per_channel']
elif self.ext == 'tif' or self.ext == 'czi':
H, W = self.reader.pages[0].shape
T = len(self.reader.pages)
return (T, H, W)
def get_frame(self, frame_idx):
"""
args
----
frame_idx : int
returns
-------
2D ndarray (YX), dtype uint16
"""
assert self._frame_valid(frame_idx)
if self.ext == 'nd2':
return self.reader.get_frame_2D(t=frame_idx)
elif self.ext == 'tif' or self.ext == 'czi':
return self.reader.pages[frame_idx].asarray()
def get_frames(self, frame_indices):
"""
args
----
frame_indices : list of int
returns
-------
3D ndarray (TYX), dtype uint16
"""
n = len(frame_indices)
result = np.empty((n, self.height, self.width),
dtype='uint16')
for i, idx in enumerate(frame_indices):
result[i,:,:] = self.get_frame(idx)
return result
def get_frame_range(self, start_frame, stop_frame):
"""
args
----
start_frame, stop_frame : int, frame indices
returns
-------
3D ndarray (TYX), dtype uint16
"""
assert self._frame_range_valid(start_frame, stop_frame)
n = stop_frame-start_frame
result = np.empty((n, self.height, self.width),
dtype='uint16')
for j, idx in enumerate(range(start_frame, stop_frame)):
result[j,:,:] = self.get_frame(idx)
return result
def sum_proj(self, start_frame=None, stop_frame=None):
"""
Return the sum intensity projection
for the full movie.
args
----
start_frame, stop_frame : int, limits
for the frames to use, if
desired
returns
-------
2D ndarray (YX), dtype uint16
"""
start_frame, stop_frame = self._process_frame_range(
start_frame, stop_frame)
result = np.zeros((self.height, self.width),
dtype='float64')
for frame_idx in range(start_frame, stop_frame):
result = result + self.get_frame(frame_idx)
return result
def max_int_proj(self, start_frame=None, stop_frame=None):
"""
Return the maximum intensity projection
for the full movie.
args
----
start_frame, stop_frame : int, limits
for the frames to use, if
desired
returns
-------
2D ndarray (YX), dtype uint16
"""
start_frame, stop_frame = self._process_frame_range(
start_frame, stop_frame)
result = np.zeros((self.height, self.width),
dtype='float64')
for frame_idx in range(start_frame, stop_frame):
result = np.maximum(result,
self.get_frame(frame_idx))
return result
def min_max(self, start_frame=None, stop_frame=None):
"""
args
----
start_frame, stop_frame : int
returns
-------
(int, int), the minimum and maximum pixel
intensities in the stack
"""
start_frame, stop_frame = self._process_frame_range(
start_frame, stop_frame)
cmax, cmin = 0, 0
for frame_idx in range(start_frame, stop_frame):
img = self.get_frame(frame_idx)
cmax = max([img.max(), cmax])
cmin = min([img.min(), cmin])
return cmin, cmax
def frame_subregion(self, frame_idx, y0=None, y1=None,
x0=None, x1=None):
"""
Return a subregion of a single frame.
args
----
frame_idx : int, the frame index
y0, y1 : int, y limits of rectangular
subregion
x0, x1 : int, x limits of rectangular
subregion
returns
-------
2D ndarray (YX), dtype uint16, the
desired subregion
"""
y0, y1, x0, x1 = self._process_subregion(y0, y1, x0, x1)
return self.get_frame(frame_idx)[y0:y1,x0:x1]
def subregion(self, y0=None, y1=None, x0=None, x1=None,
start_frame=None, stop_frame=None):
"""
Return a subregion of the movie.
args
----
y0, y1 : int, the y limits of the rectangular
subregion
x0, x1 : int, the x limits of the rectangular
subregion
start_frame, stop_frame : int, the temporal
limits on the subregion
returns
-------
3D ndarray (TYX), dtype uint16
"""
y0, y1, x0, x1 = self._process_subregion(y0, y1, x0, x1)
start_frame, stop_frame = self._process_frame_range(
start_frame, stop_frame)
T = stop_frame - start_frame
N = y1 - y0
M = x1 - x0
result = np.empty((T, N, M), dtype='uint16')
for j, frame_idx in enumerate(range(start_frame, stop_frame)):
result[j,:,:] = self.get_frame(frame_idx)[y0:y1,x0:x1]
return result
def imsave(self, out_tif, **subregion_kwargs):
"""
Save a portion of the image to a TIF
file.
args
----
out_tif : str
subregion_kwargs : to self.subregion(),
which rectangular subregion to save
returns
-------
None
"""
tifffile.imsave(
out_tif,
self.subregion(**subregion_kwargs)
)
def __init__(self, path):
self.filename = path
self.czi = CziFile(path)
def open_image(img_path, meta_path=None, z_first=False):
"""Open a multichannel 3D microscopy image.
Open a multichannel 3D microscopy image and return the image as a 4-D numpy
array. Accepted filetypes are CZI image (`.czi`) or OME-TIFF image
(`.ome.tiff` or `.ome.tif`). Also return the metadata as an XML
ElementTree object, either scraped from the CZI image or grabbed from the
XML file provided at `meta_path`. By default, the image is returned with
the dimensions [c, x, y, z].
For CZI images, the `czifile` package is required. For OME-TIFF images, the
`bioformats` and `javabridge` packages are required, as well as Java.
Parameters:
img_path: str
Path to the microscopy image. CZI and OME-TIFF files are accepted.
Optional:
meta_path: str (default None)
Path to the metadata XML file. If a `img_path` points to a CZI file,
this argument is not required, but if provided, it will be used over
the CZI metadata. If `img_path` points to an OME-TIFF file, `meta_path`
will be used to read the image's metadata. If `meta_path` is not
provided, search the image directory for an XML file with the same
basename as `img_path`.
z_first: bool (default False)
If True, return `img` with the dimensions [c, z, x, y].
"""
# determine image filetype and open
if img_path.lower().endswith('.czi'):
img_name = os.path.splitext(os.path.basename(img_path))[0]
# open CZI format using `czifile` library
from czifile import CziFile
with CziFile(img_path) as czi_file:
img_czi = czi_file.asarray()
if meta_path is not None:
metatree = ElementTree.parse(meta_path)
else:
meta = czi_file.metadata()
metatree = ElementTree.fromstring(meta)
if z_first:
img = img_czi[0, 0, :, 0, :, :, :, 0] # c, z, x, y
else:
img = img_czi[0, 0, :, 0, :, :, :, 0].transpose(0, 2, 3,
1) # c, x, y, z
elif any(
[img_path.lower().endswith(ext) for ext in ['.ome.tiff', '.ome.tif']]):
img_name = os.path.splitext(
os.path.splitext(os.path.basename(img_path))[0])[0]
# open OME-TIFF format using `bioformats` library (requires Java)
import javabridge
import bioformats
javabridge.start_vm(class_path=bioformats.JARS)
# iterate over z-stack until end of file
slices = []
for z in count():
try:
s = bioformats.load_image(img_path, z=z)
slices.append(s)
except javabridge.jutil.JavaException:
# final z-slice was read, stop reading
javabridge.kill_vm()
break
img_ome = np.stack(slices)
if z_first:
img = img_ome.transpose(3, 0, 1, 2) # c, z, x, y
else:
img = img_ome.transpose(3, 1, 2, 0) # c, x, y, z
# look for metadata .XML file with same filename
if meta_path is None:
meta_path = os.path.splitext(
os.path.splitext(img_path)[0])[0] + '.xml'
try:
metatree = ElementTree.parse(meta_path)
except IOError:
# metadata file not found
raise IOError('CZI metadata XML not found at expected path "' +
meta_path + '" (required for OME-TIFF)')
else:
raise ValueError(
'Image filetype not recognized. Allowed: .CZI, .OME.TIFF')
return img, img_name, metatree
#from scipy.misc import imresize #import matplotlib as plt from matplotlib import pylab as pl import urllib.request import os #fn='/home/pwatkins/Downloads/0.08 lead_continuous 150-50-50-03.czi' #fn = '/data/pwatkins/kara/sample_alignment_data_from_Kara_20180125/2017/training-zebrafish-4-sections/' +\ # 'zebrafish_20171013_10-15-10/003_Region3.czi' url = 'https://keeper.mpdl.mpg.de/f/ec98ecaff5674dfea904/?dl=1' fn = 'face.czi' urllib.request.urlretrieve(url, fn) czi = CziFile(fn) #print(czi.metadata, dir(czi.metadata)) iimg = czi.asarray() print(iimg.shape) os.remove(fn) #shape = np.array(iimg.shape[3:5]) #img = np.zeros(shape//10 + 1,dtype=iimg.dtype) # #for scene in range(iimg.shape[1]): # cimg = np.squeeze(iimg[0,scene,0,:,:,0])[0::10,0::10] # sel = (cimg > 0) # img[sel] = cimg[sel] img = np.squeeze(iimg)
return np.interp(image, (np.min(image), np.max(image)), (vmin, vmax))
# define arguments
parser = argparse.ArgumentParser()
parser.add_argument('image',
type=str,
nargs=1,
help='Path to .CZI image file.')
# parse arguments
args = vars(parser.parse_args())
img_path = args['image'][0]
# open CZI image
with CziFile(img_path) as image_file:
img_czi_format = image_file.asarray()
# print(image_file.axes)
# meta = image_file.metadata()
# mtree = ElementTree.fromstring(meta)
img = img_czi_format[0, 0, :, 0, 0, :, :, 0] # c, x, y
img_dapi = normalize_image(img[3, :, :])
img_fish = normalize_image(img[2, :, :])
img_dcas13 = img[1, :, :]
# fig, ax = plt.subplots(1,3)
# ax[0].imshow(img_dapi, cmap='binary_r')
# ax[0].set_title('DAPI')
# ax[1].imshow(img_fish, cmap='binary_r')
# ax[1].set_title('FISH')
def read_dtype(self, file_name: Path) -> str:
with CziFile(file_name) as czi_file:
return czi_file.dtype.name # pylint: disable=no-member
def animate_zstack_from_czi(czi, gain=1., separate=False, **kwargs):
"""
Animate the z-stack in a given CZI file. Unless separate is True, plot \
all channels as an overlay, with the legend constructed from CZI file \
metadata. If separate is True, plot all channels on separate subplots.
Currently only works with Airyscan images.
"""
def update_frame(f):
for k, imxy in enumerate(imgs_xy):
imxy.set_data(channels[k][:,:,f])
return imxy
# interpret `czi` argument
if isinstance(czi, str):
# interpret as path to CZI file
with CziFile(czi) as czi_file:
meta = czi_file.metadata()
img = czi_file.asarray()
elif isinstance(czi, CziFile):
# interpret as CziFile object
meta = czi.metadata()
img = czi.asarray()
else:
# unknown argument type
raise TypeError('`czi` argument must be of type `str` or `CziFile`.')
gain = float(gain)
# attempt to get channels and make legend
mtree = ElementTree.fromstring(meta)
track_tree = mtree.find('Metadata').find('Experiment').find('ExperimentBlocks').find('AcquisitionBlock').find('MultiTrackSetup').findall('TrackSetup')
tracks = []
for track in track_tree:
name = track.get('Name')
wavelen = float(track.find('Attenuators').find('Attenuator').find('Wavelength').text)*1E9
rgb = np.array(wavelength_to_rgb(wavelen))/256.
# print rgb
tracks.append([name, rgb])
channels = []
for t in range(len(tracks)):
channels.append(img[0, 0, t, 0, :, :, :, 0].transpose(1, 2, 0)) # x, y, z
# make colormaps
cmaps = []
for j, track in enumerate(tracks):
cmaps.append(LinearSegmentedColormap.from_list('cmap' + str(j), [list(tracks[j][1]) + [0.], list(tracks[j][1]) + [1.]]))
frames = list(range(channels[0].shape[2]))
vmins = [np.amin(channels[i]) for i in range(len(channels))]
vmaxs = [np.amax(channels[i])/gain for i in range(len(channels))]
if not separate:
fig, ax = plt.subplots()
ax.set_facecolor('k')
imgs_xy = []
for i in range(len(channels)):
imgs_xy.append(ax.imshow(channels[i][:,:,frames[0]], vmin=vmins[i], vmax=vmaxs[i], cmap=cmaps[i], **kwargs))
# ax.legend([l[0] for l in tracks])
else:
fig, ax = plt.subplots(1, len(channels))
imgs_xy = []
for i in range(len(channels)):
ax[i].set_facecolor('k')
imgs_xy.append(ax[i].imshow(channels[i][:,:,frames[0]], vmin=vmins[i], vmax=vmaxs[i], cmap=cmaps[i], **kwargs))
ax[i].set_title(tracks[i][0])
anim = FuncAnimation(fig, update_frame, frames=frames, interval=200, blit=False)
# plt.show()
return True
def read_czi(czi_path):
with CziFile(czi_path) as czi:
image_arrays = czi.asarray()
return image_arrays
import sys
sys.path.insert(0, '../../czifile/czifile/')
from czifile import CziFile
import matplotlib.pyplot as plt
import numpy as np
with CziFile('data/3500000427_100X_20170120_F05_P27.czi') as czi:
image_arrays = czi.asarray()
print(image_arrays.shape)
images = [image_arrays[0, 0, 0, idx] for idx in range(39)]
for i in range(3):
plt.figure(i + 1)
ax = plt.subplot(311)
plt.tight_layout()
ax.set_title('sample #{:d}'.format(i * 3))
ax.axis('off')
plt.imshow(np.squeeze(images[i * 3]))
ax = plt.subplot(312)
plt.tight_layout()
ax.set_title('sample #{:d}'.format(i * 3 + 1))
ax.axis('off')
plt.imshow(np.squeeze(images[i * 3 + 1]))
ax = plt.subplot(313)
plt.tight_layout()
ax.set_title('sample #{:d}'.format(i * 3 + 2))
names = []
pathfile = []
for row in czifiles.index:
names.append(czifiles.loc[row][:-4])
pathfile.append(os.path.join(targetdir, czifiles.loc[row]))
#generate a pandas dataframe listing files and paths
xlFrame = pd.DataFrame(pathfile, index=names, columns=['Pathfile'])
#save xlsfile
xlspath = os.path.join(targetdir, 'SummaryFile.xlsx')
xlFrame.to_excel(xlspath)
#load images
xlFrame['Image'] = ''
for row in xlFrame.index:
#xlFrame['Image'].loc[row] = tifffile.imread( xlFrame['Pathfile'].loc[row])
with CziFile(xlFrame['Pathfile'].loc[row]) as czi:
img = np.squeeze(czi.asarray())
img = np.rollaxis(img, 2, 0)
img = np.flipud(img)
img = np.rollaxis(img, 0, 3)
xlFrame['Image'].loc[row] = img
#normal intensity values within image
def normalizeImage(img):
img = img - np.min(img)
return img / np.max(img)
#determines the range
def setrange(x, xrange, limits):
class CziImageStack(ImageStack):
extensions = ('.czi', )
priority = 500
def open(self, location, **kwargs):
with warnings.catch_warnings():
warnings.simplefilter('ignore')
self.czi = CziFile(location.path)
self.frames = {
_get_subblock_identifier(subblock): subblock
for subblock in self.czi.filtered_subblock_directory
}
self.size = _dim_shape_to_dict(self.czi.axes, self.czi.shape)
self.metadata = self.czi.metadata(raw=False)
scaling = self.metadata['ImageDocument']['Metadata']['Scaling'][
'Items']['Distance']
# /ImageDocument
calibration_x = float(
next(iter([scale for scale in scaling if scale['Id'] == 'X'
]))['Value']) * 1E6
calibration_y = float(
next(iter([scale for scale in scaling if scale['Id'] == 'Y'
]))['Value']) * 1E6
assert calibration_x == calibration_y
self.calibration = calibration_x
timestamps = None
for entry in self.czi.attachment_directory:
if entry.name == 'TimeStamps':
timestamps = entry.data_segment().data()
break
self.timestamps = timestamps
positions = []
for scene in sorted(
self.metadata['ImageDocument']['Metadata']['Information']
['Image']['Dimensions']['S']['Scenes']['Scene'],
key=lambda scene_: int(scene_['Index'])):
x, y = scene['CenterPosition'].split(',')
positions.append((float(x), float(y)))
self.positions = positions
self.set_dimensions_and_sizes([
Dimensions.Time, Dimensions.PositionXY, Dimensions.PositionZ,
Dimensions.Channel
], [
self.size.get('T', 1),
self.size.get('S', 1), 1,
self.size.get('C', 1)
])
# noinspection PyProtectedMember
def notify_fork(self):
self.czi._fh.close()
self.czi._fh.open()
def get_data(self, what):
channel = what.get(Dimensions.Channel, 0)
position = what.get(Dimensions.PositionXY, 0)
time = what.get(Dimensions.Time, 0)
return _get_image_from_subblock(self.frames[_normalize(
_only_existing_dim(self.size, dict(C=channel, S=position,
T=time)))])
def get_meta(self, what):
try:
time = float(self.timestamps[what[Dimensions.Time]])
except TypeError:
time = what[Dimensions.Time]
try:
position = (
self.positions[what[Dimensions.PositionXY]][0],
self.positions[what[Dimensions.PositionXY]][1],
0.0,
)
except KeyError:
position = (float('nan'), float('nan'), 0.0)
meta = self.__class__.Metadata(time=time,
position=position,
calibration=self.calibration)
return meta
