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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Self-supervised masked image modeling has shown promising results on naturalimages. However, directly applying such methods to medical images remainschallenging. This difficulty stems from the complexity and distinctcharacteristics of lesions compared to natural images, which impedes effectiverepresentation learning. Additionally, conventional high fixed masking ratiosrestrict reconstructing fine lesion details, limiting the scope of learnableinformation. To tackle these limitations, we propose a novel self-supervisedmedical image segmentation framework, Adaptive Masking Lesion Patches (AMLP).Specifically, we design a Masked Patch Selection (MPS) strategy to identify andfocus learning on patches containing lesions. Lesion regions are scarce yetcritical, making their precise reconstruction vital. To reducemisclassification of lesion and background patches caused by unsupervisedclustering in MPS, we introduce an Attention Reconstruction Loss (ARL) to focuson hard-to-reconstruct patches likely depicting lesions. We further propose aCategory Consistency Loss (CCL) to refine patch categorization based onreconstruction difficulty, strengthening distinction between lesions andbackground. Moreover, we develop an Adaptive Masking Ratio (AMR) strategy thatgradually increases the masking ratio to expand reconstructible information andimprove learning. Extensive experiments on two medical segmentation datasetsdemonstrate AMLP's superior performance compared to existing self-supervisedapproaches. The proposed strategies effectively address limitations in applyingmasked modeling to medical images, tailored to capturing fine lesion detailsvital for segmentation tasks.

AMLP:Adaptive Masking Lesion Patches for Self-supervised Medical Image Segmentation