Patch Extraction

WSI.patch module

Title: ValidPath Toolbox - patch extraction module

Description: This is the patch extraction module for the ValidPath toolbox. It is includes two classes and several methods

Classes: WSIpatch_extractor, PatchExtractor

Methods: There are two methods in the patch extraction module as follows:

PatchExtractor.gen_patch(INPUTDIR: str, PatchSize: tuple, Number_of_Patches: int, intensity_check: boolean, OUTPUTDIR:str)

WSIpatch_extractor.patch_extraction(wsi_obj: object, PatchSize: tuple, OUTPUTDIR:str, Random: boolean, Visualize: boolean, Intensity_check: boolean, Number_of_Patches: (int)

class WSI.patch.PatchExtractor[source]

Bases: object

find_between(s, first, last)[source]

gen_patch(INPUTDIR, PatchSize, Number_of_Patches, intensity_check, intensity_threshold, OUTPUTDIR)[source]

This function extracts a number of pactches from extracted annotations. It can save the extracted annottions to the output directory as defined in inputs. Before running this function, please call annotation.ann_extractor.extract_ann(save_dir, XMLs, WSIs) to generate annotations. The output directory will be generated based on the strucutr of the input directories. IF the WSI Magnification is 13X or 20X, this code will automaticall convert to 20X.

Parameters:

INPUTDIRstring: the path to the input directory
PatchSizetuple: the size of image patches to be extracted
Number_of_Patchesint: the number of patches per annotation to be extracted
intensity_checkboolean: to filter the image patches and eliminate empty ones

OUTPUTDIR : string the path to the output directory to save image patches

Returns:

Image – extracted image patches from the annotated area.

class WSI.patch.WSIpatch_extractor[source]

Bases: object

patch_extraction(patch_size, output_folder, random_state, visualize, intensity_check, intensity_threshold, std_threshold, patch_number=-1)[source]

this function Generate object for tiles using the DeepZoomGenerator and divided the svs file into tiles of size 256 with no overlap. then processing and saving each tile to local directory.

Parameters:

wsi_objobject: an object containing WSI file and its information
patch_size: integer: the size of image patches to be extracted
output_folderstring: the path to the output directory to save image patches
random_stateboolean: extract patches randomly or in order
visualize: boolean: either to plot extracted patches or not
intensity_check: boolean: to filter the image patches and eliminate empty ones
intensity_threshold: integer: the threshold to include image patches
std_threshold: integer: the standard deviation threhold to include image patches
patch_numberboolean: the number of patches to be extracted. Set to ‘-1’ to extract all possible image patches

patch_extraction_of_tissue(patch_size, output_folder, number_of_patches=1, vis=False)[source]

patch_extraction_with_normalized_tiles(patch_size, output_folder, random_state=True, patch_number=-1)[source]

this function Generate object for tiles using the DeepZoomGenerator and divided the svs file into tiles of size 256 with no overlap. then processing and saving each tile to local directory.

Parameters:

wsi_objobject: WSI object.
patch_size: integer: size tiles
output_folderstring: path root folder to save tiles

perform_segmentation_state: boolean random_state : boolean