Periphery-aware COVID-19 diagnosis with contrastive representation enhancement

Highlights：

• A novel diagnosis approach with spatial pattern prior and representation enhancement mechanism is proposed to distinguish COVID-19 in the complex scenario of multi-type pneumonia classification.

• An important spatial pattern prior is introduced into the deep network by learning a Periphery-aware Spatial Prediction (PSP) task.

• An adaptive Contrastive Representation Enhancement (CRE) mechanism is designed to effectively capture the intra-class similarity and inter-class difference of various types of pneumonia.

• A large-scale chest CT dataset with 3D and 2D samples of four categories is collected for both volume-level and slice-level diagnosis research.

摘要

•A novel diagnosis approach with spatial pattern prior and representation enhancement mechanism is proposed to distinguish COVID-19 in the complex scenario of multi-type pneumonia classification.•An important spatial pattern prior is introduced into the deep network by learning a Periphery-aware Spatial Prediction (PSP) task.•An adaptive Contrastive Representation Enhancement (CRE) mechanism is designed to effectively capture the intra-class similarity and inter-class difference of various types of pneumonia.•A large-scale chest CT dataset with 3D and 2D samples of four categories is collected for both volume-level and slice-level diagnosis research.

论文关键词：Automated COVID-19 diagnosis,Chest CT images,Periphery-aware spatial prediction (PSP),Contrastive representation enhancement (CRE)

论文评审过程：Received 4 January 2021, Revised 25 March 2021, Accepted 27 April 2021, Available online 6 May 2021, Version of Record 18 May 2021.