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

For medical image segmentation, contrastive learning is the dominant practiceto improve the quality of visual representations by contrasting semanticallysimilar and dissimilar pairs of samples. This is enabled by the observationthat without accessing ground truth labels, negative examples with trulydissimilar anatomical features, if sampled, can significantly improve theperformance. In reality, however, these samples may come from similaranatomical regions and the models may struggle to distinguish the minoritytail-class samples, making the tail classes more prone to misclassification,both of which typically lead to model collapse. In this paper, we propose ARCO,a semi-supervised contrastive learning (CL) framework with stratified grouptheory for medical image segmentation. In particular, we first propose buildingARCO through the concept of variance-reduced estimation and show that certainvariance-reduction techniques are particularly beneficial in pixel/voxel-levelsegmentation tasks with extremely limited labels. Furthermore, we theoreticallyprove these sampling techniques are universal in variance reduction. Finally,we experimentally validate our approaches on eight benchmarks, i.e., five 2D/3Dmedical and three semantic segmentation datasets, with different labelsettings, and our methods consistently outperform state-of-the-artsemi-supervised methods. Additionally, we augment the CL frameworks with thesesampling techniques and demonstrate significant gains over previous methods. Webelieve our work is an important step towards semi-supervised medical imagesegmentation by quantifying the limitation of current self-supervisionobjectives for accomplishing such challenging safety-critical tasks.

Rethinking Semi-Supervised Medical Image Segmentation: A Variance-Reduction Perspective