wsipipe.load.annotations package

All annotations are loaded in the generic annotation format. Individual modules convert specific annotation types to the generic

annotation module

Parent classes that contain functionality for reading annotations. These are used to render different types of annotations into a common format

class Annotation(name, annotation_type, label, vertices)

Bases: object

Class for a single annotation.

There can be multiple annotations on a slide

Parameters:

• name (str) – Name of the annotation.

• type (str) – One of Dot, Polygon, Spline or Rectangle

• label (str) – What label should be given to the annotation

• vertices (List[PointF]) – A list of vertices, each of which is an PointF object, a named tuple (x, y) of floats.

• annotation_type (str) –

draw(image, labels, factor)

Renders the annotation into the image.

Parameters:

• image (np.array) – Array to write the annotations into, must have dtype float.

• labels (Dict[str, int]) – The value to write into the image for each type of label.

• factor (float) – How much to scale (by divison) each vertex by.

class AnnotationSet(annotations, labels, labels_order, fill_label)

Bases: object

Class for all annotations on a slide.

Parameters:

• annotations (List[Annotation]) – A list of all Annotations on a slide

• labels (Dict[str, int]) – A dictionary where the keys are the names of labels, with the integer values with which the string should be replaced.

• labels_order (List[str]) – An order the labels should be plotted in. Where annotations overlap they will be drawn in this order, so the final label will be on top

• fill_label (str) – The label given to any unannotated areas.

render(shape, factor)

Creates a labelled image containing annotations

This creates an array of size = shape, that is factor times smaller than the level at which the annotation vertexes are specified. Annotations vertex positions are assumed to be specified at level 0, and therefore for many WSI a np.array of the same size as level 0 would not fit in memory. Therefore one factor times smaller is created.

Parameters:

• shape (Shape) – size of numpy array to create

• factor (float) – How much to scale (by divison) each vertex by.

Return type:

numpy.array

visualise_annotations(annot_path, slide_path, loader, level, slide_label)

Creates a image render of the annotations of a slide

Converts annotations from level zero to the specified level. Requires slide path to find the correct dimensions of the output image. Returns a numpy array

Parameters:

• annot_path (Path) – A path to the annotation file

• slide_path (Path) – A path to the WSI file

• loader – The loader to use for slides and annots

• level (int) – the level to create the numpy array

• slide_label (str) –

Returns:

An array the same size as the WSI at level with the annotation labels plotted in it.

Return type:

labels_image (np.array)

asapxml module

Functions to load annotations stored in asapxml formats and convert to Annotation class formats

annotation_from_tag(tag, group_labels)

Convert an asapxml element to annotation format.

Parameters:

• tag (Element) – An element from the xml Element tree

• group_labels (Dict[str, str]) – A dictionary of group labels that convert values stored in xml PartOfGroup to required label. e.g {“Tumor”: “tumor”, “Metastasis”: “tumor”, “Normal”: “normal”, “Tissue”: “normal”}

Return type:

Annotation

load_annotations_asapxml(xml_file_path, group_labels)

Read xml file and create annotations

Parameters:

• xml_file_path (Path) – PAth to xml file to read

• group_labels (Dict[str, str]) – A dictionary of group labels that convert values stored in xml PartOfGroup (keys) to required label (values). e.g {“Tumor”: “tumor”, “Metastasis”: “tumor”, “Normal”: “normal”, “Tissue”: “normal”}

Return type:

List[Annotation]