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

Biomedical imaging datasets are often small and biased, meaning thatreal-world performance of predictive models can be substantially lower thanexpected from internal testing. This work proposes using generative imageediting to simulate dataset shifts and diagnose failure modes of biomedicalvision models; this can be used in advance of deployment to assess readiness,potentially reducing cost and patient harm. Existing editing methods canproduce undesirable changes, with spurious correlations learned due to theco-occurrence of disease and treatment interventions, limiting practicalapplicability. To address this, we train a text-to-image diffusion model onmultiple chest X-ray datasets and introduce a new editing method RadEdit thatuses multiple masks, if present, to constrain changes and ensure consistency inthe edited images. We consider three types of dataset shifts: acquisitionshift, manifestation shift, and population shift, and demonstrate that ourapproach can diagnose failures and quantify model robustness without additionaldata collection, complementing more qualitative tools for explainable AI.

RadEdit: stress-testing biomedical vision models via diffusion image editing