introvoyz041's picture
Migrated from GitHub
12d2e9e verified
"""
Copyright 2021, Dana-Farber Cancer Institute and Weill Cornell Medicine
License: GNU GPL 2.0
"""
import os
import reprlib
from pathlib import Path
import anndata
import dask.distributed
import h5py
import matplotlib.pyplot as plt
import numpy as np
from loguru import logger
import pathml.core
import pathml.preprocessing.pipeline
from pathml.core.h5managers import h5pathManager
from pathml.core.slide_types import SlideType
def infer_backend(path):
"""
Checks file extensions to try to infer correct backend to use.
Uses the file extensions from the sets contained in this file (pathml/core/slide_data.py)
For file formats which are supported by both openslide and bioformats, will return "bioformats".
Args:
path: path to file
Returns:
str: one of "bioformats", "openslide", "dicom", "h5path"
"""
path = str(path)
for extension_set, name in zip(
[pathmlext, bioformatsext, openslideext, dicomext],
["h5path", "bioformats", "openslide", "dicom"],
):
for ext in extension_set:
if path[-len(ext) :] == ext:
return name
raise ValueError(
f"input path {path} doesn't match any supported file extensions"
) # pragma: no cover
class SlideData:
"""
Main class representing a slide and its annotations.
Args:
filepath (str): Path to file on disk.
name (str, optional): name of slide. If ``None``, and a ``filepath`` is provided, name defaults to filepath.
masks (pathml.core.Masks, optional): object containing {key, mask} pairs
tiles (pathml.core.Tiles, optional): object containing {coordinates, tile} pairs
labels (collections.OrderedDict, optional): dictionary containing {key, label} pairs
backend (str, optional): backend to use for interfacing with slide on disk.
Must be one of {"OpenSlide", "BioFormats", "DICOM", "h5path"} (case-insensitive).
Note that for supported image formats, OpenSlide performance can be significantly better than BioFormats.
Consider specifying ``backend = "openslide"`` when possible.
If ``None``, and a ``filepath`` is provided, tries to infer the correct backend from the file extension.
Defaults to ``None``.
slide_type (pathml.core.SlideType, optional): slide type specification. Must be a
:class:`~pathml.core.SlideType` object. Alternatively, slide type can be specified by using the
parameters ``stain``, ``tma``, ``rgb``, ``volumetric``, and ``time_series``.
stain (str, optional): Flag indicating type of slide stain. Must be one of [‘HE’, ‘IHC’, ‘Fluor’].
Defaults to ``None``. Ignored if ``slide_type`` is specified.
platform (str, optional): Flag indicating the imaging platform (e.g. CODEX, Vectra, etc.).
Defaults to ``None``. Ignored if ``slide_type`` is specified.
tma (bool, optional): Flag indicating whether the image is a tissue microarray (TMA).
Defaults to ``False``. Ignored if ``slide_type`` is specified.
rgb (bool, optional): Flag indicating whether the image is in RGB color.
Defaults to ``None``. Ignored if ``slide_type`` is specified.
volumetric (bool, optional): Flag indicating whether the image is volumetric.
Defaults to ``None``. Ignored if ``slide_type`` is specified.
time_series (bool, optional): Flag indicating whether the image is a time series.
Defaults to ``None``. Ignored if ``slide_type`` is specified.
counts (anndata.AnnData): object containing counts matrix associated with image quantification
"""
def __init__(
self,
filepath,
name=None,
masks=None,
tiles=None,
labels=None,
backend=None,
slide_type=None,
stain=None,
platform=None,
tma=None,
rgb=None,
volumetric=None,
time_series=None,
counts=None,
dtype=None,
):
# check inputs
assert masks is None or isinstance(
masks, dict
), f"mask are of type {type(masks)} but must be type dict"
if labels:
assert all(
[isinstance(key, str) for key in labels.keys()]
), f"Input label keys are of types {[type(k) for k in labels.keys()]}. All label keys must be of type str."
assert all(
[
isinstance(val, (str, np.ndarray))
or np.issubdtype(type(val), np.number)
or np.issubdtype(type(val), np.bool_)
for val in labels.values()
]
), (
f"Input label vals are of types {[type(v) for v in labels.values()]}. "
f"All label values must be of type str or np.ndarray or a number (i.e. a subdtype of np.number) "
)
assert tiles is None or (
isinstance(tiles, list)
and all([isinstance(tile, pathml.core.Tile) for tile in tiles])
), f"tiles are of type {type(tiles)} but must be a list of objects of type pathml.core.tiles.Tile"
assert slide_type is None or isinstance(
slide_type, pathml.core.SlideType
), f"slide_type is of type {type(slide_type)} but must be of type pathml.core.types.SlideType"
assert backend is None or (
isinstance(backend, str)
and backend.lower() in {"openslide", "bioformats", "dicom", "h5path"}
), f"backend {backend} must be one of ['OpenSlide', 'BioFormats', 'DICOM', 'h5path'] (case-insensitive)."
assert counts is None or isinstance(
counts, anndata.AnnData
), f"counts is if type {type(counts)} but must be of type anndata.AnnData"
# instantiate SlideType object if needed
if not slide_type and any([stain, platform, tma, rgb, volumetric, time_series]):
stain_type_dict = {
"stain": stain,
"platform": platform,
"tma": tma,
"rgb": rgb,
"volumetric": volumetric,
"time_series": time_series,
}
# remove any Nones
stain_type_dict = {
key: val for key, val in stain_type_dict.items() if val is not None
}
if stain_type_dict:
slide_type = pathml.core.slide_types.SlideType(**stain_type_dict)
# get name from filepath if no name is provided
if name is None and filepath is not None:
name = Path(filepath).name
_load_from_h5path = False
if backend:
# convert everything to lower so it's case insensitive
backend = backend.lower()
else:
# try to infer the correct backend
backend = infer_backend(filepath)
if backend == "h5path":
_load_from_h5path = True
if backend.lower() == "openslide":
backend_obj = pathml.core.OpenSlideBackend(filepath)
elif backend.lower() == "bioformats":
backend_obj = pathml.core.BioFormatsBackend(filepath, dtype)
elif backend.lower() == "dicom":
backend_obj = pathml.core.DICOMBackend(filepath)
elif backend.lower() == "h5path":
backend_obj = None
else:
raise ValueError(f"invalid backend: {repr(backend)}.")
self._filepath = filepath if filepath else None
self.backend = backend
self.slide = backend_obj if backend_obj else None
self.name = name
self.labels = labels
self.slide_type = slide_type
if _load_from_h5path:
# populate the SlideData object from existing h5path file
with h5py.File(filepath, "r") as f:
self.h5manager = h5pathManager(h5path=f)
self.name = self.h5manager.h5["fields"].attrs["name"]
self.labels = {
key: val
for key, val in self.h5manager.h5["fields"]["labels"].attrs.items()
}
# empty dict evaluates to False
if not self.labels:
self.labels = None
slide_type = {
key: val
for key, val in self.h5manager.h5["fields"]["slide_type"].attrs.items()
if val is not None
}
if slide_type:
self.slide_type = SlideType(**slide_type)
else:
self.h5manager = h5pathManager(slidedata=self)
self.masks = pathml.core.Masks(h5manager=self.h5manager, masks=masks)
self.tiles = pathml.core.Tiles(h5manager=self.h5manager, tiles=tiles)
def __repr__(self):
out = []
out.append(f"SlideData(name={repr(self.name)}")
out.append(f"slide_type={repr(self.slide_type)}")
if self._filepath:
out.append(f"filepath='{self._filepath}'")
if self.backend:
out.append(f"backend={repr(self.backend)}")
out.append(f"image shape: {self.shape}")
try:
nlevels = self.slide.level_count
# TODO: change to specific exception
except Exception:
nlevels = 1
out.append(f"number of levels: {nlevels}")
out.append(repr(self.tiles))
out.append(repr(self.masks))
if self.tiles:
out.append(f"tile_shape={eval(self.tiles.tile_shape)}")
if self.labels:
out.append(
f"{len(self.labels)} labels: {reprlib.repr(list(self.labels.keys()))}"
)
else:
out.append("labels=None")
if self.counts:
out.append(f"counts matrix of shape {self.counts.shape}")
else:
out.append("counts=None")
out = ",\n\t".join(out)
out += ")"
return out
def run(
self,
pipeline,
distributed=True,
client=None,
tile_size=256,
tile_stride=None,
level=0,
tile_pad=False,
overwrite_existing_tiles=False,
write_dir=None,
**kwargs,
):
"""
Run a preprocessing pipeline on SlideData.
Tiles are generated by calling self.generate_tiles() and pipeline is applied to each tile.
Args:
pipeline (pathml.preprocessing.pipeline.Pipeline): Preprocessing pipeline.
distributed (bool): Whether to distribute model using client. Defaults to True.
client: dask.distributed client
tile_size (int, optional): Size of each tile. Defaults to 256px
tile_stride (int, optional): Stride between tiles. If ``None``, uses ``tile_stride = tile_size``
for non-overlapping tiles. Defaults to ``None``.
level (int, optional): Level to extract tiles from. Defaults to ``None``.
tile_pad (bool): How to handle chunks on the edges. If ``True``, these edge chunks will be zero-padded
symmetrically and yielded with the other chunks. If ``False``, incomplete edge chunks will be ignored.
Defaults to ``False``.
overwrite_existing_tiles (bool): Whether to overwrite existing tiles. If ``False``, running a pipeline will
fail if ``tiles is not None``. Defaults to ``False``.
write_dir (str): Path to directory to write the processed slide to. The processed SlideData object
will be written to the directory immediately after the pipeline has completed running.
The filepath will default to "<write_dir>/<slide.name>.h5path. Defaults to ``None``.
**kwargs: Other arguments passed through to ``generate_tiles()`` method of the backend.
"""
assert isinstance(
pipeline, pathml.preprocessing.pipeline.Pipeline
), f"pipeline is of type {type(pipeline)} but must be of type pathml.preprocessing.pipeline.Pipeline"
assert self.slide is not None, "cannot run pipeline because self.slide is None"
if len(self.tiles) != 0:
# in this case, tiles already exist
if not overwrite_existing_tiles:
raise Exception(
"Slide already has tiles. Running the pipeline will overwrite the existing tiles. Use overwrite_existing_tiles=True to force overwriting existing tiles."
)
else:
# delete all existing tiles
for tile_key in self.tiles.keys:
self.tiles.remove(tile_key)
# TODO: be careful here since we are modifying h5 outside of h5manager
# look into whether we can push this into h5manager
if tile_stride is None:
tile_stride = tile_size
elif isinstance(tile_stride, int):
tile_stride = (tile_stride, tile_stride)
self.h5manager.h5["tiles"].attrs["tile_stride"] = tile_stride
shutdown_after = False
if distributed:
if client is None:
client = dask.distributed.Client()
shutdown_after = True
logger.info(
f"creating a default distributed.Client(): {client.scheduler_info()}"
)
# map pipeline application onto each tile
processed_tile_futures = []
for tile in self.generate_tiles(
level=level,
shape=tile_size,
stride=tile_stride,
pad=tile_pad,
**kwargs,
):
if not tile.slide_type:
tile.slide_type = self.slide_type
# explicitly scatter data, i.e. send the tile data out to the cluster before applying the pipeline
# according to dask, this can reduce scheduler burden and keep data on workers
big_future = client.scatter(tile)
f = client.submit(pipeline.apply, big_future)
processed_tile_futures.append(f)
# as tiles are processed, add them to h5
for future, tile in dask.distributed.as_completed(
processed_tile_futures, with_results=True
):
self.tiles.add(tile)
if shutdown_after:
client.shutdown()
else:
for tile in self.generate_tiles(
level=level,
shape=tile_size,
stride=tile_stride,
pad=tile_pad,
**kwargs,
):
if not tile.slide_type:
tile.slide_type = self.slide_type
pipeline.apply(tile)
self.tiles.add(tile)
if write_dir:
self.write(Path(write_dir) / f"{self.name}.h5path")
@property
def shape(self):
"""
Convenience method for getting the image shape.
Calling ``wsi.shape`` is equivalent to calling ``wsi.slide.get_image_shape()`` with default arguments.
Returns:
Tuple[int, int]: Shape of image (H, W)
"""
if self.backend == "h5path":
return tuple(self.h5manager.h5["fields"].attrs["shape"])
else:
return self.slide.get_image_shape()
def extract_region(self, location, size, *args, **kwargs):
"""
Extract a region of the image.
This is a convenience method which passes arguments through to the ``extract_region()`` method of whichever
backend is in use. Refer to documentation for each backend.
Args:
location (Tuple[int, int]): Location of top-left corner of tile (i, j)
size (Union[int, Tuple[int, int]]): Size of each tile. May be a tuple of (height, width) or a
single integer, in which case square tiles of that size are generated.
*args: positional arguments passed through to ``extract_region()`` method of the backend.
**kwargs: keyword arguments passed through to ``extract_region()`` method of the backend.
Returns:
np.ndarray: image at the specified region
"""
if self.slide is None: # pragma: no cover
raise ValueError(
"Cannot call `.extract_region()` because no slide is specified. "
"If already tiled, access `.tiles` directly instead"
)
return np.squeeze(self.slide.extract_region(location, size, *args, **kwargs))
def generate_tiles(self, shape=3000, stride=None, pad=False, **kwargs):
"""
Generator over Tile objects containing regions of the image.
Calls ``generate_tiles()`` method of the backend.
Tries to add the corresponding slide-level masks to each tile, if possible.
Adds slide-level labels to each tile, if possible.
Args:
shape (int or tuple(int)): Size of each tile. May be a tuple of (height, width) or a single integer,
in which case square tiles of that size are generated. Defaults to 256px.
stride (int): stride between chunks. If ``None``, uses ``stride = size`` for non-overlapping chunks.
Defaults to ``None``.
pad (bool): How to handle tiles on the edges. If ``True``, these edge tiles will be zero-padded
and yielded with the other chunks. If ``False``, incomplete edge chunks will be ignored.
Defaults to ``False``.
**kwargs: Other arguments passed through to ``generate_tiles()`` method of the backend.
Yields:
pathml.core.tile.Tile: Extracted Tile object
"""
for tile in self.slide.generate_tiles(shape, stride, pad, **kwargs):
# add masks for tile, if possible
# i.e. if the SlideData has a Masks object, and the tile has coordinates
if self.masks is not None and tile.coords is not None:
# masks not supported if pad=True
# to implement, need to update Mask.slice to support slices that go beyond the full mask
if not pad:
i, j = tile.coords
di, dj = tile.image.shape[0:2]
# add the Masks object for the masks corresponding to the tile
# this assumes that the tile didn't already have any masks
# this should work since the backend reads from image only
# adding safety check just in case to make sure we don't overwrite any existing mask
# if this assertion fails, we will need to rewrite this part
assert (
len(tile.masks) == 0
), "tile yielded from backend already has mask. slide_data.generate_tiles is trying to overwrite it"
tile_slices = [slice(i, i + di), slice(j, j + dj)]
tile.masks = self.masks.slice(tile_slices)
# add slide-level labels to each tile, if possible
if self.labels is not None:
tile.labels = self.labels
# add slidetype to tile
if tile.slide_type is None:
tile.slide_type = self.slide_type
yield tile
def plot(self, ax=None):
"""
View a thumbnail of the image, using matplotlib.
Not supported by all backends.
Args:
ax: matplotlib axis object on which to plot the thumbnail. Optional.
"""
try:
thumbnail = self.slide.get_thumbnail(size=(500, 500))
# TODO: change to specific exception
except Exception:
if not self.slide:
raise NotImplementedError(
"Plotting only supported via backend, but SlideData has no backend."
)
else:
raise NotImplementedError(
f"plotting not supported for slide_backend={self.slide.__class__.__name__}"
)
if ax is None:
ax = plt.gca()
ax.imshow(thumbnail)
if self.name:
ax.set_title(self.name)
ax.axis("off")
@property
def counts(self):
return self.tiles.h5manager.counts if self.tiles.h5manager else None
@counts.setter
def counts(self, value):
if self.tiles.h5manager:
assert value is None or isinstance(
value, anndata.AnnData
), f"cannot set counts with obj of type {type(value)}. Must be Anndata"
self.tiles.h5manager.counts = value
else:
raise AttributeError(
"cannot assign counts slidedata contains no tiles, first generate tiles"
)
def write(self, path):
"""
Write contents to disk in h5path format.
Args:
path (Union[str, bytes, os.PathLike]): path to file to be written
"""
path = Path(path)
pathdir = Path(os.path.dirname(path))
pathdir.mkdir(parents=True, exist_ok=True)
with h5py.File(path, "w") as f:
for ds in self.h5manager.h5.keys():
self.h5manager.h5.copy(ds, f)
if self.counts:
pathml.core.utils.writecounts(f["counts"], self.counts)
class HESlide(SlideData):
"""
Convenience class to load a SlideData object for H&E slides.
Passes through all arguments to ``SlideData()``, along with ``slide_type = types.HE`` flag.
Refer to :class:`~pathml.core.slide_data.SlideData` for full documentation.
"""
def __init__(self, *args, **kwargs):
kwargs["slide_type"] = pathml.core.types.HE
super().__init__(*args, **kwargs)
class MultiparametricSlide(SlideData):
"""
Convenience class to load a SlideData object for multiparametric immunofluorescence slides.
Passes through all arguments to ``SlideData()``, along with ``slide_type = types.IF`` flag and default ``backend = "bioformats"``.
Refer to :class:`~pathml.core.slide_data.SlideData` for full documentation.
"""
def __init__(self, *args, **kwargs):
kwargs["slide_type"] = pathml.core.types.IF
if "backend" not in kwargs:
kwargs["backend"] = "bioformats"
super().__init__(*args, **kwargs)
class IHCSlide(SlideData):
"""
Convenience class to load a SlideData object for IHC slides.
Passes through all arguments to ``SlideData()``, along with ``slide_type = types.IHC`` flag.
Refer to :class:`~pathml.core.slide_data.SlideData` for full documentation.
"""
def __init__(self, *args, **kwargs):
kwargs["slide_type"] = pathml.core.types.IHC
super().__init__(*args, **kwargs)
class VectraSlide(SlideData):
"""
Convenience class to load a SlideData object for Vectra (Polaris) slides.
Passes through all arguments to ``SlideData()``, along with ``slide_type = types.Vectra`` flag and default ``backend = "bioformats"``.
Refer to :class:`~pathml.core.slide_data.SlideData` for full documentation.
"""
def __init__(self, *args, **kwargs):
kwargs["slide_type"] = pathml.core.types.Vectra
if "backend" not in kwargs:
kwargs["backend"] = "bioformats"
super().__init__(*args, **kwargs)
class CODEXSlide(SlideData):
"""
Convenience class to load a SlideData object from Akoya Biosciences CODEX format.
Passes through all arguments to ``SlideData()``, along with ``slide_type = types.CODEX`` flag and default ``backend = "bioformats"``.
Refer to :class:`~pathml.core.slide_data.SlideData` for full documentation.
# TODO:
hierarchical biaxial gating (flow-style analysis)
"""
def __init__(self, *args, **kwargs):
kwargs["slide_type"] = pathml.core.types.CODEX
if "backend" not in kwargs:
kwargs["backend"] = "bioformats"
super().__init__(*args, **kwargs)
# dicts used to infer correct backend from file extension
pathmlext = {".h5", ".h5path"}
openslideext = {
".svs",
".tif",
".tiff",
".ndpi",
".vms",
".vmu",
".scn",
".mrxs",
".svslide",
".bif",
}
bioformatsext = {
".tiff",
".tif",
".sld",
".aim",
".al3d",
".gel",
".am",
".amiramesh",
".grey",
".hx",
".labels",
".cif",
".img",
".hdr",
".sif",
".png",
".afi",
".htd",
".pnl",
".avi",
".arf",
".exp",
".spc",
".sdt",
".xml",
".1sc",
".pic",
".raw",
".scn",
".ims",
".img",
".cr2",
".crw",
".ch5",
".c01",
".dib",
".vsi",
".wpi",
".dv",
".r3d",
".rcpnl",
".eps",
".epsi",
".ps",
".fits",
".dm3",
".dm4",
".dm2",
".vff",
".naf",
".his",
".i2i",
".ics",
".ids",
".fff",
".seq",
".ipw",
".hed",
".mod",
".liff",
".obf",
".msr",
".xdce",
".frm",
".inr",
".hdr",
".ipl",
".ipm",
".dat",
".par",
".jp2",
".j2k",
".jpf",
".jpk",
".jpx",
".klb",
".xv",
".bip",
".fli",
".msr",
".lei",
".lif",
".scn",
".sxm",
".l2d",
".lim",
".stk",
".nd",
".htd",
".mnc",
".mrw",
".mng",
".stp",
".mrc",
".st",
".ali",
".map",
".rec",
".mrcs",
".nef",
".hdr",
".nii",
".nii.gz",
".nrrd",
".nhdr",
".apl",
".mtb",
".tnb",
".obsep",
".oib",
".oif",
".oir",
".ome.tiff",
".ome.tif",
".ome.tf2",
".ome.tf8",
".ome.btf",
".ome.xml",
".ome",
".top",
".pcoraw",
".rec",
".pcx",
".pds",
".im3",
".qptiff",
".pbm",
".pgm",
".ppm",
".psd",
".bin",
".pict",
".cfg",
".spe",
".afm",
".mov",
".sm2",
".sm3",
".xqd",
".xqf",
".cxd",
".spi",
".stk",
".tga",
".db",
".vws",
".tfr",
".ffr",
".zfr",
".zfp",
".2fl",
".sld",
".pr3",
".dat",
".hdr",
".fdf",
".bif",
".dti",
".xys",
".mvd2",
".acff",
".wat",
".bmp",
".wlz",
".lms",
".zvi",
".czi",
".lsm",
".mdb",
}
dicomext = {".dicom", ".dcm"}