Multi-task contrastive learning for automatic CT and X-ray diagnosis of COVID-19

作者：

Highlights：

• A novel Contrastive Multi-Task Convolutional Neural Network (CMT-CNN) is proposed for automatic COVID-19 diagnosis.

• The main task is to diagnose COVID-19 from other pneumonia and normal controls. The auxiliary task is self-supervised contrastive learning to acquire transformation-invariant representations.

• A series of interpretable transformations are defined for medical image augmentation.

• Extensive experiments demonstrate that the auxiliary task can significantly improve the generalization of CNN on both CT and X-ray datasets.

摘要

•A novel Contrastive Multi-Task Convolutional Neural Network (CMT-CNN) is proposed for automatic COVID-19 diagnosis.•The main task is to diagnose COVID-19 from other pneumonia and normal controls. The auxiliary task is self-supervised contrastive learning to acquire transformation-invariant representations.•A series of interpretable transformations are defined for medical image augmentation.•Extensive experiments demonstrate that the auxiliary task can significantly improve the generalization of CNN on both CT and X-ray datasets.

论文关键词：Computed tomography,X-ray,COVID-19,Deep learning,Multi-task learning,Contrastive learning

论文评审过程：Received 20 June 2020, Revised 18 January 2021, Accepted 24 January 2021, Available online 26 January 2021, Version of Record 12 February 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107848