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

Today, deep convolutional neural networks (CNNs) have demonstratedstate-of-the-art performance for medical image segmentation, on various imagingmodalities and tasks. Despite early success, segmentation networks may stillgenerate anatomically aberrant segmentations, with holes or inaccuracies nearthe object boundaries. To enforce anatomical plausibility, recent researchstudies have focused on incorporating prior knowledge such as object shape orboundary, as constraints in the loss function. Prior integrated could below-level referring to reformulated representations extracted from theground-truth segmentations, or high-level representing external medicalinformation such as the organ's shape or size. Over the past few years,prior-based losses exhibited a rising interest in the research field since theyallow integration of expert knowledge while still being architecture-agnostic.However, given the diversity of prior-based losses on different medical imagingchallenges and tasks, it has become hard to identify what loss works best forwhich dataset. In this paper, we establish a benchmark of recent prior-basedlosses for medical image segmentation. The main objective is to provideintuition onto which losses to choose given a particular task or dataset. Tothis end, four low-level and high-level prior-based losses are selected. Theconsidered losses are validated on 8 different datasets from a variety ofmedical image segmentation challenges including the Decathlon, the ISLES andthe WMH challenge. Results show that whereas low-level prior-based losses canguarantee an increase in performance over the Dice loss baseline regardless ofthe dataset characteristics, high-level prior-based losses can increaseanatomical plausibility as per data characteristics.

Effect of Prior-based Losses on Segmentation Performance: A Benchmark