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

While making a tremendous impact in various fields, deep neural networksusually require large amounts of labeled data for training which are expensiveto collect in many applications, especially in the medical domain. Unlabeleddata, on the other hand, is much more abundant. Semi-supervised learningtechniques, such as co-training, could provide a powerful tool to leverageunlabeled data. In this paper, we propose a novel framework, uncertainty-awaremulti-view co-training (UMCT), to address semi-supervised learning on 3D data,such as volumetric data from medical imaging. In our work, co-training isachieved by exploiting multi-viewpoint consistency of 3D data. We generatedifferent views by rotating or permuting the 3D data and utilize asymmetrical3D kernels to encourage diversified features in different sub-networks. Inaddition, we propose an uncertainty-weighted label fusion mechanism to estimatethe reliability of each view's prediction with Bayesian deep learning. As oneview requires the supervision from other views in co-training, ourself-adaptive approach computes a confidence score for the prediction of eachunlabeled sample in order to assign a reliable pseudo label. Thus, our approachcan take advantage of unlabeled data during training. We show the effectivenessof our proposed semi-supervised method on several public datasets from medicalimage segmentation tasks (NIH pancreas & LiTS liver tumor dataset). Meanwhile,a fully-supervised method based on our approach achieved state-of-the-artperformances on both the LiTS liver tumor segmentation and the MedicalSegmentation Decathlon (MSD) challenge, demonstrating the robustness and valueof our framework, even when fully supervised training is feasible.

3D Semi-Supervised Learning with Uncertainty-Aware Multi-View Co-Training