MID-UNet: Multi-input directional UNet for COVID-19 lung infection segmentation from CT images

作者：

Highlights：

• We propose a multidimensional input to reflect the GGOs or infiltration features.

• We propose a directional convolution block to represent the fibrotic-streak features.

• We propose a novel region contour loss function to restric irregular boundaries.

摘要

•We propose a multidimensional input to reflect the GGOs or infiltration features.•We propose a directional convolution block to represent the fibrotic-streak features.•We propose a novel region contour loss function to restric irregular boundaries.

论文关键词：COVID-19,Infection segmentation,CT image,Deep learning,Convolutional neural networks

论文评审过程：Received 28 July 2021, Revised 30 May 2022, Accepted 23 July 2022, Available online 2 August 2022, Version of Record 13 August 2022.

论文官网地址：https://doi.org/10.1016/j.image.2022.116835