English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In pre-clinical pathology, there is a paradox between the abundance of rawdata (whole slide images from many organs of many individual animals) and thelack of pixel-level slide annotations done by pathologists. Due to timeconstraints and requirements from regulatory authorities, diagnoses are insteadstored as slide labels. Weakly supervised training is designed to takeadvantage of those data, and the trained models can be used by pathologists torank slides by their probability of containing a given lesion of interest. Inthis work, we propose a novel contextualized eXplainable AI (XAI) framework andits application to deep learning models trained on Whole Slide Images (WSIs) inDigital Pathology. Specifically, we apply our methods to amulti-instance-learning (MIL) model, which is trained solely on slide-levellabels, without the need for pixel-level annotations. We validatequantitatively our methods by quantifying the agreements of our explanations'heatmaps with pathologists' annotations, as well as with predictions from asegmentation model trained on such annotations. We demonstrate the stability ofthe explanations with respect to input shifts, and the fidelity with respect toincreased model performance. We quantitatively evaluate the correlation betweenavailable pixel-wise annotations and explainability heatmaps. We show that theexplanations on important tiles of the whole slide correlate with tissuechanges between healthy regions and lesions, but do not exactly behave like ahuman annotator. This result is coherent with the model training strategy.

From slides (through tiles) to pixels: an explainability framework for weakly supervised models in pre-clinical pathology