Multi-organ segmentation network for abdominal CT images based on spatial attention and deformable convolution

作者：

Highlights：

• A novel network based on U-Net for accurate abdominal multi-organ segmentation.

• Spatial attention map is learned though external supervision to highlight organs.

• Deformable convolution to handle variable shapes and sizes.

• Improve skip-connection structure by combining multi-scale attention maps.

摘要

•A novel network based on U-Net for accurate abdominal multi-organ segmentation.•Spatial attention map is learned though external supervision to highlight organs.•Deformable convolution to handle variable shapes and sizes.•Improve skip-connection structure by combining multi-scale attention maps.

论文关键词：Deep learning,Multi-organ segmentation,U-net,Attention mechanism,Deformable convolution,CT image

论文评审过程：Received 30 September 2021, Revised 25 July 2022, Accepted 16 August 2022, Available online 19 August 2022, Version of Record 3 September 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118625