Progressive global perception and local polishing network for lung infection segmentation of COVID-19 CT images
作者：
Highlights：
• This paper aims to segment COVID-19-caused lung infection regions from CT images.
• A progressive global perception and local polishing deep network was constructed.
• We aggregated the multi-scale and multi-level features in a coarse-to-fine fashion.
• We trained the encoder-decoder network by a boundary-aware multiple supervision way.
• Our results can intuitively show the severity of disease to help doctors.
摘要
•This paper aims to segment COVID-19-caused lung infection regions from CT images.•A progressive global perception and local polishing deep network was constructed.•We aggregated the multi-scale and multi-level features in a coarse-to-fine fashion.•We trained the encoder-decoder network by a boundary-aware multiple supervision way.•Our results can intuitively show the severity of disease to help doctors.
论文关键词：Coronavirus disease 2019 (COVID-19),Global perception,Local polishing,Feature recursive aggregation,Multiple supervision
论文评审过程：Received 1 April 2021, Revised 23 June 2021, Accepted 6 July 2021, Available online 11 July 2021, Version of Record 24 July 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108168
