Contour-enhanced attention CNN for CT-based COVID-19 segmentation

作者：

Highlights：

• Accurate detection of COVID-19 is one of the challenging research topics in today's healthcare sector to control the coronavirus pandemic.

• The proposed Attention Decoder CNN fuses shape, boundary information from CT contours to enhance discriminability of COVID-19 infection.

• A pixel-precise attention upsampling module has been proposed to leverage encoder context towards reconstructing the high-resolution segmentation map.

• The proposed CNN displayed state-of-the-art results with a high 85.43% dice score and 88.10% recall on the combined MosMed and Jun Ma dataset.

摘要

•Accurate detection of COVID-19 is one of the challenging research topics in today's healthcare sector to control the coronavirus pandemic.•The proposed Attention Decoder CNN fuses shape, boundary information from CT contours to enhance discriminability of COVID-19 infection.•A pixel-precise attention upsampling module has been proposed to leverage encoder context towards reconstructing the high-resolution segmentation map.•The proposed CNN displayed state-of-the-art results with a high 85.43% dice score and 88.10% recall on the combined MosMed and Jun Ma dataset.

论文关键词：COVID-19,Segmentation,Deep learning,Attention,Decoder, CNN

论文评审过程：Received 4 September 2020, Revised 14 September 2021, Accepted 14 January 2022, Available online 19 January 2022, Version of Record 24 January 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108538