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

Medical imaging AI systems such as disease classification and segmentationare increasingly inspired and transformed from computer vision based AIsystems. Although an array of adversarial training and/or loss function baseddefense techniques have been developed and proved to be effective in computervision, defending against adversarial attacks on medical images remains largelyan uncharted territory due to the following unique challenges: 1) labelscarcity in medical images significantly limits adversarial generalizability ofthe AI system; 2) vastly similar and dominant fore- and background in medicalimages make it hard samples for learning the discriminating features betweendifferent disease classes; and 3) crafted adversarial noises added to theentire medical image as opposed to the focused organ target can make clean andadversarial examples more discriminate than that between different diseaseclasses. In this paper, we propose a novel robust medical imaging AI frameworkbased on Semi-Supervised Adversarial Training (SSAT) and UnsupervisedAdversarial Detection (UAD), followed by designing a new measure for assessingsystems adversarial risk. We systematically demonstrate the advantages of ourrobust medical imaging AI system over the existing adversarial defensetechniques under diverse real-world settings of adversarial attacks using abenchmark OCT imaging data set.

Defending against adversarial attacks on medical imaging AI system, classification or detection?