MetaCOVID: A Siamese neural network framework with contrastive loss for n-shot diagnosis of COVID-19 patients

作者：

Highlights：

• A Siamese neural network framework for COVID-19 diagnosis from CXR images.

• Benefit of using contrastive loss and n-shot learning in design of the framework.

• A fine-tuned pre-trained CNN encoder to capture unbiased feature representations.

• The diagnosis problem is formulated as a k-way n-shot classification problem.

• Experimental results with a limited dataset to show efficacy of the framework.

摘要

•A Siamese neural network framework for COVID-19 diagnosis from CXR images.•Benefit of using contrastive loss and n-shot learning in design of the framework.•A fine-tuned pre-trained CNN encoder to capture unbiased feature representations.•The diagnosis problem is formulated as a k-way n-shot classification problem.•Experimental results with a limited dataset to show efficacy of the framework.

论文关键词：COVID-19 diagnosis,Multi-shot learning,Contrastive loss,CXR images,Siamese network

论文评审过程：Received 3 August 2020, Revised 23 September 2020, Accepted 13 October 2020, Available online 17 October 2020, Version of Record 19 February 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107700