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

Unsupervised domain adaptation (UDA) has been a vital protocol for migratinginformation learned from a labeled source domain to facilitate theimplementation in an unlabeled heterogeneous target domain. Although UDA istypically jointly trained on data from both domains, accessing the labeledsource domain data is often restricted, due to concerns over patient dataprivacy or intellectual property. To sidestep this, we propose "off-the-shelf(OS)" UDA (OSUDA), aimed at image segmentation, by adapting an OS segmentortrained in a source domain to a target domain, in the absence of source domaindata in adaptation. Toward this goal, we aim to develop a novel batch-wisenormalization (BN) statistics adaptation framework. In particular, we graduallyadapt the domain-specific low-order BN statistics, e.g., mean and variance,through an exponential momentum decay strategy, while explicitly enforcing theconsistency of the domain shareable high-order BN statistics, e.g., scaling andshifting factors, via our optimization objective. We also adaptively quantifythe channel-wise transferability to gauge the importance of each channel, viaboth low-order statistics divergence and a scaling factor.~Furthermore, weincorporate unsupervised self-entropy minimization into our framework to boostperformance alongside a novel queued, memory-consistent self-training strategyto utilize the reliable pseudo label for stable and efficient unsupervisedadaptation. We evaluated our OSUDA-based framework on both cross-modality andcross-subtype brain tumor segmentation and cardiac MR to CT segmentation tasks.Our experimental results showed that our memory consistent OSUDA performsbetter than existing source-relaxed UDA methods and yields similar performanceto UDA methods with source data.

Memory Consistent Unsupervised Off-the-Shelf Model Adaptation for Source-Relaxed Medical Image Segmentation