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/

Zendy Plus

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

Zendy Tools

Zendy Open

Deep Learning failure cases are abundant, particularly in the medical area.Recent studies in out-of-distribution generalization have advanced considerablyon well-controlled synthetic datasets, but they do not represent medicalimaging contexts. We propose a pipeline that relies on artifacts annotation toenable generalization evaluation and debiasing for the challenging skin lesionanalysis context. First, we partition the data into levels of increasinglyhigher biased training and test sets for better generalization assessment.Then, we create environments based on skin lesion artifacts to enable domaingeneralization methods. Finally, after robust training, we perform a test-timedebiasing procedure, reducing spurious features in inference images. Ourexperiments show our pipeline improves performance metrics in biased cases, andavoids artifacts when using explanation methods. Still, when evaluating suchmodels in out-of-distribution data, they did not prefer clinically-meaningfulfeatures. Instead, performance only improved in test sets that present similarartifacts from training, suggesting models learned to ignore the known set ofartifacts. Our results raise a concern that debiasing models towards a singleaspect may not be enough for fair skin lesion analysis.

Artifact-Based Domain Generalization of Skin Lesion Models