Collaborative boundary-aware context encoding networks for error map prediction

作者：

Highlights：

• We explore the network structure design for error map segmentation with image-prediction as input.

• A novel collaborative boundary-aware context encoding framework called EP-Net is proposed for error map segmentation.

• The proposed EP-Net achieves strong performance on error segmentation task, which can accurately localize the small and thin error regions.

• The segmentation quality assessment based on the error segmentation results is more reliable and has stronger transferability compared with other methods.

摘要

•We explore the network structure design for error map segmentation with image-prediction as input.•A novel collaborative boundary-aware context encoding framework called EP-Net is proposed for error map segmentation.•The proposed EP-Net achieves strong performance on error segmentation task, which can accurately localize the small and thin error regions.•The segmentation quality assessment based on the error segmentation results is more reliable and has stronger transferability compared with other methods.

论文关键词：Segmentation quality assessment,Error map prediction,Medical image segmentation

论文评审过程：Received 17 February 2021, Revised 3 August 2021, Accepted 28 December 2021, Available online 31 December 2021, Version of Record 6 January 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108515