GraphXCOVID: Explainable deep graph diffusion pseudo-Labelling for identifying COVID-19 on chest X-rays

作者：

Highlights：

• A new deep semi-supervised framework for identifying COVID-19 in ChestX-ray.

• A novel optimisation model for psedo-label generation based on the p=1 Dirichlet energy.

• Our technique reports high sensitivity in COVID-19 and performance requiring minimal labels.

• We provide meaningful visualisation to the radiologist for judging whether the diagnosis is correct.

摘要

•A new deep semi-supervised framework for identifying COVID-19 in ChestX-ray.•A novel optimisation model for psedo-label generation based on the p=1 Dirichlet energy.•Our technique reports high sensitivity in COVID-19 and performance requiring minimal labels.•We provide meaningful visualisation to the radiologist for judging whether the diagnosis is correct.

论文关键词：COVID-19,Chest X-ray,Semi-Supervised learning,Deep learning,Explainability

论文评审过程：Received 7 December 2020, Revised 20 August 2021, Accepted 21 August 2021, Available online 26 August 2021, Version of Record 4 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108274