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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Training a deep neural network is an optimization problem with four mainingredients: the design of the deep neural network, the per-sample lossfunction, the population loss function, and the optimizer. However, methodsdeveloped to compete in recent BraTS challenges tend to focus only on thedesign of deep neural network architectures, while paying less attention to thethree other aspects. In this paper, we experimented with adopting the oppositeapproach. We stuck to a generic and state-of-the-art 3D U-Net architecture andexperimented with a non-standard per-sample loss function, the generalizedWasserstein Dice loss, a non-standard population loss function, correspondingto distributionally robust optimization, and a non-standard optimizer, Ranger.Those variations were selected specifically for the problem of multi-classbrain tumor segmentation. The generalized Wasserstein Dice loss is a per-sampleloss function that allows taking advantage of the hierarchical structure of thetumor regions labeled in BraTS. Distributionally robust optimization is ageneralization of empirical risk minimization that accounts for the presence ofunderrepresented subdomains in the training dataset. Ranger is a generalizationof the widely used Adam optimizer that is more stable with small batch size andnoisy labels. We found that each of those variations of the optimization ofdeep neural networks for brain tumor segmentation leads to improvements interms of Dice scores and Hausdorff distances. With an ensemble of three deepneural networks trained with various optimization procedures, we achievedpromising results on the validation dataset of the BraTS 2020 challenge. Ourensemble ranked fourth out of the 693 registered teams for the segmentationtask of the BraTS 2020 challenge.

Generalized Wasserstein Dice Score, Distributionally Robust Deep Learning, and Ranger for brain tumor segmentation: BraTS 2020 challenge