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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Magnetic resonance (MR) protocols rely on several sequences to assesspathology and organ status properly. Despite advances in image analysis, wetend to treat each sequence, here termed modality, in isolation. Takingadvantage of the common information shared between modalities (an organ'sanatomy) is beneficial for multi-modality processing and learning. However, wemust overcome inherent anatomical misregistrations and disparities in signalintensity across the modalities to obtain this benefit. We present a methodthat offers improved segmentation accuracy of the modality of interest (over asingle input model), by learning to leverage information present in othermodalities, even if few (semi-supervised) or no (unsupervised) annotations areavailable for this specific modality. Core to our method is learning adisentangled decomposition into anatomical and imaging factors. Sharedanatomical factors from the different inputs are jointly processed and fused toextract more accurate segmentation masks. Image misregistrations are correctedwith a Spatial Transformer Network, which non-linearly aligns the anatomicalfactors. The imaging factor captures signal intensity characteristics acrossdifferent modality data and is used for image reconstruction, enablingsemi-supervised learning. Temporal and slice pairing between inputs are learneddynamically. We demonstrate applications in Late Gadolinium Enhanced (LGE) andBlood Oxygenation Level Dependent (BOLD) cardiac segmentation, as well as in T2abdominal segmentation. Code is available athttps://github.com/vios-s/multimodal_segmentation.

Disentangle, Align and Fuse for Multimodal and Semi-Supervised Image Segmentation