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

The scarcity of labeled data often impedes the application of deep learningto the segmentation of medical images. Semi-supervised learning seeks toovercome this limitation by exploiting unlabeled examples in the learningprocess. In this paper, we present a novel semi-supervised segmentation methodthat leverages mutual information (MI) on categorical distributions to achieveboth global representation invariance and local smoothness. In this method, wemaximize the MI for intermediate feature embeddings that are taken from boththe encoder and decoder of a segmentation network. We first propose a global MIloss constraining the encoder to learn an image representation that isinvariant to geometric transformations. Instead of resorting tocomputationally-expensive techniques for estimating the MI on continuousfeature embeddings, we use projection heads to map them to a discrete clusterassignment where MI can be computed efficiently. Our method also includes alocal MI loss to promote spatial consistency in the feature maps of the decoderand provide a smoother segmentation. Since mutual information does not requirea strict ordering of clusters in two different assignments, we incorporate afinal consistency regularization loss on the output which helps align thecluster labels throughout the network. We evaluate the method on fourchallenging publicly-available datasets for medical image segmentation.Experimental results show our method to outperform recently-proposed approachesfor semi-supervised segmentation and provide an accuracy near to fullsupervision while training with very few annotated images.

Boosting Semi-supervised Image Segmentation with Global and Local Mutual Information Regularization