eduzhai > Applied Sciences > Engineering >

CVR-Net A deep convolutional neural network for coronavirus recognition from chest radiography images

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 31 pages

Abstract: The novel Coronavirus Disease 2019 (COVID-19) is a global pandemic diseasespreading rapidly around the world. A robust and automatic early recognition ofCOVID-19, via auxiliary computer-aided diagnostic tools, is essential fordisease cure and control. The chest radiography images, such as ComputedTomography (CT) and X-ray, and deep Convolutional Neural Networks (CNNs), canbe a significant and useful material for designing such tools. However,designing such an automated tool is challenging as a massive number of manuallyannotated datasets are not publicly available yet, which is the corerequirement of supervised learning systems. In this article, we propose arobust CNN-based network, called CVR-Net (Coronavirus Recognition Network), forthe automatic recognition of the coronavirus from CT or X-ray images. Theproposed end-to-end CVR-Net is a multi-scale-multi-encoder ensemble model,where we have aggregated the outputs from two different encoders and theirdifferent scales to obtain the final prediction probability. We train and testthe proposed CVR-Net on three different datasets, where the images havecollected from different open-source repositories. We compare our proposedCVR-Net with state-of-the-art methods, which are trained and tested on the samedatasets. We split three datasets into five different tasks, where each taskhas a different number of classes, to evaluate the multi-tasking CVR-Net. Ourmodel achieves an overall F1-score & accuracy of 0.997 & 0.998; 0.963 & 0.964;0.816 & 0.820; 0.961 & 0.961; and 0.780 & 0.780, respectively, for task-1 totask-5. As the CVR-Net provides promising results on the small datasets, it canbe an auspicious computer-aided diagnostic tool for the diagnosis ofcoronavirus to assist the clinical practitioners and radiologists. Our sourcecodes and model are publicly available atthis https URL.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...