patches¶

errors¶

filters¶

A module to filter patches in a slide.

exception Error¶

Bases: Exception

Base of custom errors.

exception UnknownMethodError¶

Bases: Error

Raise when no class is found in a datafolder.

standardize_filters(filters, top_level, low_level)¶

Check validity of hierarchical filters.

Parameters:

filters (Union[str, Sequence[Sequence[Union[str, Callable]]], Dict[int, Sequence[Sequence[Union[str, Callable]]]]]) – filters to apply. Can be formatted as a single string, a list or a dictionary mapping a level to corresponding filters.
top_level (int) – top pyramid level to consider.
low_level (int) – lowest pyramid level to consider.

Return type:

Dict[int, Sequence[Sequence[Union[str, Callable]]]]

Returns:

Dictionnary mapping each level to corresponding filters.

filter_hasdapi(image, dapi_channel=0, tolerance=1)¶

Give presence of dapi in a patch.

Parameters:

image (uint8)) – image of a patch.
dapi_channel (int) – channel to extract dapi signal.
tolerance (int) – value on dapi intensity encountered to accept a patch.

Return type:

bool

Returns:

Whether dapi is visible in slide.

filter_has_significant_dapi(image, dapi_channel=0, tolerance=0.5, dapi_tolerance=1)¶

Give if enough dapi is present in a patch.

Parameters:

image (uint8)) – image of a patch.
dapi_channel (int) – channel to extract dapi signal.
tolerance (float) – part of the patch that must contain dapi.
dapi_tolerance (int) – value on dapi intensity encountered to accept a patch.

Return type:

bool

Returns:

Whether dapi is significantally visible in slide.

get_tissue_from_rgb(image, blacktol=0, whitetol=230)¶

Return the tissue mask segmentation of an image.

True pixels for the tissue, false pixels for the background.

Parameters:

image (uint8)) – image of a patch.
blacktol (Union[float, int]) – tolerance value for black pixels.
whitetol (Union[float, int]) – tolerance value for white pixels.

Return type:

bool8)

Returns:

Mask where true pixels are tissue, false are background.

get_tissue_from_lab(image, blacktol=5, whitetol=90)¶

Get the tissue mask segmentation of an image.

This version operates in the lab space, conversion of the image from rgb to lab is performed first.

Parameters:

image (uint8)) – image of a patch.
blacktol (Union[float, int]) – tolerance value for black pixels.
whitetol (Union[float, int]) – tolerance value for white pixels.

Return type:

bool8)

Returns:

Mask where true pixels are tissue, false are background.

get_tissue(image, blacktol=5, whitetol=90, method='lab')¶

Get the tissue mask segmentation of an image.

One can choose the segmentation method.

Parameters:

image (uint8)) – image of a patch.
blacktol (Union[float, int]) – tolerance value for black pixels.
whitetol (Union[float, int]) – tolerance value for white pixels.
method (str) – one of ‘lab’ or ‘rgb’, function to be called.

Return type:

bool8)

Returns:

Mask where true pixels are tissue, false are background.

filter_has_tissue_he(image, blacktol=5, whitetol=90)¶

Return true if tissue inside the patch.

Filters tissue using the l channel from lab space. Conversion of the image from rgb to lab is performed first.

Parameters:

image (uint8)) – image of a patch.
blacktol (Union[float, int]) – tolerance value for black pixels.
whitetol (Union[float, int]) – tolerance value for white pixels.

Return type:

bool

Returns:

True if tissue is detected.

functional_api¶

object_api¶

slide_filters¶

filter_remove_small_objects(in_mask, avoid_overmask=True, overmask_thresh=10, min_size_fac=1)¶

Removes small objects from a binary mask. Can recursively lowers its minimum accepted size if too much tissue is erased.

Parameters:

in_mask (bool8)) – input binary mask (must be boolean).
avoid_overmask (bool) – if True recursively call itself if too much tissue is erased.
overmask_thresh (float) – if avoid_overmask and more than overmask_thresh% of the input mask is erased, calls the function recursively with lower minimum accepted size.
min_size_fac (float) – multiplier for overmask_thresh used to compute minimum accepted size. Mainly for internal use in recursive call.

Return type:

bool8)

Returns:

Output binary mask with small objects removed.

filter_thumbnail(x)¶

Compute a tissue mask from a slide thumbnail.

Filters background, red pen, blue pen using La*b* space.

Parameters:: x (uint8)) – input thumbnail as a numpy byte array.
Return type:: bool8)
Returns:: Numpy binary mask where usable tissue is marked as True.

filter_fluo_thumbnail(x, channel=2)¶

Compute a tissue mask from a fluorescent slide thumbnail.

Filters empty regions using otsu operator.

Parameters:

x (uint8)) – input thumbnail as a numpy byte array.
channel (int) – fluorescent channel to perform threshold on, default is 2 for blue <=> dapi.

Return type:

bool8)

Returns:

Numpy binary mask where usable tissue is marked as True.

get_json2pathaia_filter(annotfile, label)¶

Get a slide_filter corresponding to annotations in Micromap format for a specific class. This filter will only keep patches that have the corresponding label.

Parameters:

annotfile (Union[str, PathLike]) – Path to annotation file for this slide.
label (str) – annotation label that is supposed to be positive.

Returns:

Numpy binary mask where annotated tissue is marked as True.

visu¶

Useful functions for visualizing patches in WSIs.

preview_from_queries(slide, queries, min_res=512, color=(255, 255, 0), thickness=2, cell_size=20, size_0=None)¶

Give thumbnail with patches displayed.

Parameters:

slide (OpenSlide) – openslide object
queries (Sequence[Patch]) – patch objects to preview from
min_res (int) – minimum size for the smallest side of the thumbnail (usually the width)
color (Tuple[int, int, int]) – rgb color for patch boundaries
thickness (int) – thickness of patch boundaries
cell_size (int) – size of a cell representing a patch in the grid
psize – size of a patch at level 0

Return type:

uint8)

Returns:

Thumbnail image with patches displayed.