You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 

364 lines
15 KiB

# -*- coding: utf-8 -*-
import struct
import array
import six
import warnings
from pims.base_frames import Frame
import numpy as np
from nd2reader.common import get_version, read_chunk
from nd2reader.exceptions import InvalidVersionError
from nd2reader.label_map import LabelMap
from nd2reader.raw_metadata import RawMetadata
class Parser(object):
"""Parses ND2 files and creates a Metadata and driver object.
"""
CHUNK_HEADER = 0xabeceda
CHUNK_MAP_START = six.b("ND2 FILEMAP SIGNATURE NAME 0001!")
CHUNK_MAP_END = six.b("ND2 CHUNK MAP SIGNATURE 0000001!")
supported_file_versions = {(3, None): True}
def __init__(self, fh):
self._fh = fh
self._label_map = None
self._raw_metadata = None
self.metadata = None
# First check the file version
self.supported = self._check_version_supported()
# Parse the metadata
self._parse_metadata()
def calculate_image_properties(self, index):
"""Calculate FOV, channels and z_levels
Args:
index(int): the index (which is simply the order in which the image was acquired)
Returns:
tuple: tuple of the field of view, the channel and the z level
"""
field_of_view = self._calculate_field_of_view(index)
channel = self._calculate_channel(index)
z_level = self._calculate_z_level(index)
return field_of_view, channel, z_level
def get_image(self, index):
"""
Creates an Image object and adds its metadata, based on the index (which is simply the order in which the image
was acquired). May return None if the ND2 contains multiple channels and not all were taken in each cycle (for
example, if you take bright field images every minute, and GFP images every five minutes, there will be some
indexes that do not contain an image. The reason for this is complicated, but suffice it to say that we hope to
eliminate this possibility in future releases. For now, you'll need to check if your image is None if you're
doing anything out of the ordinary.
Args:
index(int): the index (which is simply the order in which the image was acquired)
Returns:
Frame: the image
"""
field_of_view, channel, z_level = self.calculate_image_properties(index)
channel_offset = index % len(self.metadata["channels"])
image_group_number = int(index / len(self.metadata["channels"]))
frame_number = self._calculate_frame_number(image_group_number, field_of_view, z_level)
try:
timestamp, image = self._get_raw_image_data(image_group_number, channel_offset, self.metadata["height"],
self.metadata["width"])
except (TypeError):
return Frame([], frame_no=frame_number, metadata=self._get_frame_metadata())
else:
return Frame(image, frame_no=frame_number, metadata=self._get_frame_metadata())
def get_image_by_attributes(self, frame_number, field_of_view, channel, z_level, height, width):
"""Gets an image based on its attributes alone
Args:
frame_number: the frame number
field_of_view: the field of view
channel_name: the color channel name
z_level: the z level
height: the height of the image
width: the width of the image
Returns:
Frame: the requested image
"""
frame_number = 0 if frame_number is None else frame_number
field_of_view = 0 if field_of_view is None else field_of_view
channel = 0 if channel is None else channel
z_level = 0 if z_level is None else z_level
image_group_number = self._calculate_image_group_number(frame_number, field_of_view, z_level)
try:
timestamp, raw_image_data = self._get_raw_image_data(image_group_number, channel,
height, width)
except (TypeError):
return Frame([], frame_no=frame_number, metadata=self._get_frame_metadata())
else:
return Frame(raw_image_data, frame_no=frame_number, metadata=self._get_frame_metadata())
@staticmethod
def get_dtype_from_metadata():
"""Determine the data type from the metadata.
For now, always use float64 to prevent unexpected overflow errors when manipulating the data (calculating sums/
means/etc.)
"""
return np.float64
def _check_version_supported(self):
"""Checks if the ND2 file version is supported by this reader.
Returns:
bool: True on supported
"""
major_version, minor_version = get_version(self._fh)
supported = self.supported_file_versions.get(
(major_version, minor_version)) or self.supported_file_versions.get((major_version, None))
if not supported:
print("Warning: No parser is available for your current ND2 version (%d.%d). " % (
major_version, minor_version) + "This might lead to unexpected behaviour.")
return supported
def _parse_metadata(self):
"""Reads all metadata and instantiates the Metadata object.
"""
# Retrieve raw metadata from the label mapping
self._label_map = self._build_label_map()
self._raw_metadata = RawMetadata(self._fh, self._label_map)
self.metadata = self._raw_metadata.__dict__
self.acquisition_times = self._raw_metadata.acquisition_times
def _build_label_map(self):
"""
Every label ends with an exclamation point, however, we can't directly search for those to find all the labels
as some of the bytes contain the value 33, which is the ASCII code for "!". So we iteratively find each label,
grab the subsequent data (always 16 bytes long), advance to the next label and repeat.
Returns:
LabelMap: the computed label map
"""
# go 8 bytes back from file end
self._fh.seek(-8, 2)
chunk_map_start_location = struct.unpack("Q", self._fh.read(8))[0]
self._fh.seek(chunk_map_start_location)
raw_text = self._fh.read(-1)
return LabelMap(raw_text)
def _calculate_field_of_view(self, index):
"""Determines what field of view was being imaged for a given image.
Args:
index(int): the index (which is simply the order in which the image was acquired)
Returns:
int: the field of view
"""
images_per_cycle = len(self.metadata["z_levels"]) * len(self.metadata["channels"])
return int((index - (index % images_per_cycle)) / images_per_cycle) % len(self.metadata["fields_of_view"])
def _calculate_channel(self, index):
"""Determines what channel a particular image is.
Args:
index(int): the index (which is simply the order in which the image was acquired)
Returns:
string: the name of the color channel
"""
return self.metadata["channels"][index % len(self.metadata["channels"])]
def _calculate_z_level(self, index):
"""Determines the plane in the z-axis a given image was taken in.
In the future, this will be replaced with the actual offset in micrometers.
Args:
index(int): the index (which is simply the order in which the image was acquired)
Returns:
The z level
"""
return self.metadata["z_levels"][int(
((index - (index % len(self.metadata["channels"]))) / len(self.metadata["channels"])) % len(
self.metadata["z_levels"]))]
def _calculate_image_group_number(self, frame_number, fov, z_level):
"""
Images are grouped together if they share the same time index, field of view, and z-level.
Args:
frame_number: the time index
fov: the field of view number
z_level: the z level number
Returns:
int: the image group number
"""
z_length = len(self.metadata['z_levels'])
z_length = z_length if z_length > 0 else 1
fields_of_view = len(self.metadata["fields_of_view"])
fields_of_view = fields_of_view if fields_of_view > 0 else 1
return frame_number * fields_of_view * z_length + (fov * z_length + z_level)
def _calculate_frame_number(self, image_group_number, field_of_view, z_level):
"""
Images are in the same frame if they share the same group number and field of view and are taken sequentially.
Args:
image_group_number: the image group number (see _calculate_image_group_number)
field_of_view: the field of view number
z_level: the z level number
Returns:
"""
return (image_group_number - (field_of_view * len(self.metadata["z_levels"]) + z_level)) / (
len(self.metadata["fields_of_view"]) * len(self.metadata["z_levels"]))
@property
def _channel_offset(self):
"""
Image data is interleaved for each image set. That is, if there are four images in a set, the first image
will consist of pixels 1, 5, 9, etc, the second will be pixels 2, 6, 10, and so forth.
Returns:
dict: the channel offset for each channel
"""
return {channel: n for n, channel in enumerate(self.metadata["channels"])}
def _get_unwanted_bytes_ids(
self, image_group_data, image_data_start, height, width
):
# Check if the byte array size conforms to the image axes size. If not, check
# that the number of unexpected (unwanted) bytes is a multiple of the number of
# rows (height), as the same unmber of unwanted bytes is expected to be
# appended at the end of each row. Then, returns the indexes of the unwanted
# bytes.
number_of_true_channels = int(len(image_group_data[4:]) / (height * width))
n_unwanted_bytes = (len(image_group_data[image_data_start:])) % (height * width)
if not n_unwanted_bytes:
return np.arange(0)
assert 0 == n_unwanted_bytes % height, (
"An unexpected number of extra bytes was encountered based on the expected"
+ " frame size, therefore the file could not be parsed."
)
return np.arange(
image_data_start + height * number_of_true_channels,
len(image_group_data) - n_unwanted_bytes + 1,
height * number_of_true_channels,
)
def _remove_bytes_by_id(self, byte_ids, image_group_data, height):
# Remove bytes by ID.
bytes_per_row = len(byte_ids) // height
warnings.warn(
f"{len(byte_ids)} ({bytes_per_row}*{height}) unexpected zero "
+ "bytes were found in the ND2 file and removed to allow further parsing."
)
for i in range(len(byte_ids)):
del image_group_data[byte_ids[i] : (byte_ids[i] + bytes_per_row)]
def _get_raw_image_data(self, image_group_number, channel_offset, height, width):
"""Reads the raw bytes and the timestamp of an image.
Args:
image_group_number: the image group number (see _calculate_image_group_number)
channel_offset: the number of the color channel
height: the height of the image
width: the width of the image
Returns:
"""
chunk = self._label_map.get_image_data_location(image_group_number)
data = read_chunk(self._fh, chunk)
# All images in the same image group share the same timestamp! So if you have complicated image data,
# your timestamps may not be entirely accurate. Practically speaking though, they'll only be off by a few
# seconds unless you're doing something super weird.
timestamp = struct.unpack("d", data[:8])[0]
image_group_data = array.array("H", data)
image_data_start = 4 + channel_offset
# Stitched ND2 files have been reported to contain unexpected (according to
# image shape) zero bytes at the end of each image data row. This hinders
# proper reshaping of the data. Hence, here the unwanted zero bytes are
# identified and removed.
unwanted_byte_ids = self._get_unwanted_bytes_ids(
image_group_data, image_data_start, height, width
)
if 0 != len(unwanted_byte_ids):
assert np.all(
image_group_data[unwanted_byte_ids + np.arange(len(unwanted_byte_ids))]
== 0
), (
f"{len(unwanted_byte_ids)} unexpected non-zero bytes were found"
+ " in the ND2 file, the file could not be parsed."
)
self._remove_bytes_by_id(unwanted_byte_ids, image_group_data, height)
# The images for the various channels are interleaved within the same array. For example, the second image
# of a four image group will be composed of bytes 2, 6, 10, etc. If you understand why someone would design
# a data structure that way, please send the author of this library a message.
number_of_true_channels = int(len(image_group_data[4:]) / (height * width))
try:
image_data = np.reshape(
image_group_data[image_data_start::number_of_true_channels],
(height, width),
)
except ValueError:
image_data = np.reshape(
image_group_data[image_data_start::number_of_true_channels],
(
height,
len(image_group_data[image_data_start::number_of_true_channels])
// height,
),
)
# Skip images that are all zeros! This is important, since NIS Elements creates blank "gap" images if you
# don't have the same number of images each cycle. We discovered this because we only took GFP images every
# other cycle to reduce phototoxicity, but NIS Elements still allocated memory as if we were going to take
# them every cycle.
if np.any(image_data):
return timestamp, image_data
# If a blank "gap" image is encountered, generate an array of corresponding height and width to avoid
# errors with ND2-files with missing frames. Array is filled with nan to reflect that data is missing.
else:
empty_frame = np.full((height, width), np.nan)
warnings.warn(
"ND2 file contains gap frames which are represented by np.nan-filled"
+ " arrays; to convert to zeros use e.g. np.nan_to_num(array)"
)
return timestamp, image_data
def _get_frame_metadata(self):
"""Get the metadata for one frame
Returns:
dict: a dictionary containing the parsed metadata
"""
return self.metadata