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

Artificial intelligence and machine learning techniques have the promise torevolutionize the field of digital pathology. However, these models demandconsiderable amounts of data, while the availability of unbiased training datais limited. Synthetic images can augment existing datasets, to improve andvalidate AI algorithms. Yet, controlling the exact distribution of cellularfeatures within them is still challenging. One of the solutions is harnessingconditional generative adversarial networks that take a semantic mask as aninput rather than a random noise. Unlike other domains, outlining the exactcellular structure of tissues is hard, and most of the input masks depictregions of cell types. However, using polygon-based masks introduce inherentartifacts within the synthetic images - due to the mismatch between the polygonsize and the single-cell size. In this work, we show that introducing randomsingle-pixel noise with the appropriate spatial frequency into a polygonsemantic mask can dramatically improve the quality of the synthetic images. Weused our platform to generate synthetic images of immunohistochemistry-treatedlung biopsies. We test the quality of the images using a three-fold validationprocedure. First, we show that adding the appropriate noise frequency yields87% of the similarity metrics improvement that is obtained by adding the actualsingle-cell features. Second, we show that the synthetic images pass the Turingtest. Finally, we show that adding these synthetic images to the train setimproves AI performance in terms of PD-L1 semantic segmentation performances.Our work suggests a simple and powerful approach for generating synthetic dataon demand to unbias limited datasets to improve the algorithms' accuracy andvalidate their robustness.

Between Generating Noise and Generating Images: Noise in the Correct  Frequency Improves the Quality of Synthetic Histopathology Images for Digital  Pathology