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

Recent advances in deep learning have led to breakthroughs in the developmentof automated skin disease classification. As we observe an increasing interestin these models in the dermatology space, it is crucial to address aspects suchas the robustness towards input data distribution shifts. Current skin diseasemodels could make incorrect inferences for test samples from different hardwaredevices and clinical settings or unknown disease samples, which areout-of-distribution (OOD) from the training samples. To this end, we propose asimple yet effective approach that detect these OOD samples prior to making anydecision. The detection is performed via scanning in the latent spacerepresentation (e.g., activations of the inner layers of any pre-trained skindisease classifier). The input samples could also perturbed to maximisedivergence of OOD samples. We validate our ODD detection approach in two usecases: 1) identify samples collected from different protocols, and 2) detectsamples from unknown disease classes. Additionally, we evaluate the performanceof the proposed approach and compare it with other state-of-the-art methods.Furthermore, data-driven dermatology applications may deepen the disparity inclinical care across racial and ethnic groups since most datasets are reportedto suffer from bias in skin tone distribution. Therefore, we also evaluate thefairness of these OOD detection methods across different skin tones. Ourexperiments resulted in competitive performance across multiple datasets indetecting OOD samples, which could be used (in the future) to design moreeffective transfer learning techniques prior to inferring on these samples.

Out-of-Distribution Detection in Dermatology using Input Perturbation and Subset Scanning