Skip to content

COVID19 detection from Radiographs: Is Deep Learning able to handle the crisis ?

License

Notifications You must be signed in to change notification settings

saeed-anwar/COVID19-Baselines

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

18 Commits
 
 
 
 
 
 

Repository files navigation

COVID19 detection from Radiographs: Is Deep Learning able to handle the crisis?

This repository is for COVID19 detection from Radiographs introduced in the following paper

Muhammad Saqib, Saeed Anwar, Abbas Anwar, Lars Petersson, Nabin Sharma, Michael Blumenstein, "COVID19 detection from Radiographs: Is Deep Learning able to handle the crisis? ", preprint, 2020

The models are built in PyTorch 1.5.1 and tested on Ubuntu 14.04/16.04 environment (Python3.8, CUDA10.2).

Contents

  1. Introduction
  2. Network
  3. Test
  4. Results
  5. Citation

Introduction

The COVID-19 is a highly contagious viral infection which played havoc on everyone's life in many different ways. According to the world health organization and scientists, more testing potentially helps governments and disease control organizations in containing the spread of the virus. The use of chest radiographs is one of the early screening tests to determine the onset of disease, as the infection affects the lungs severely. This study will investigate and automate the process of testing by using state-of-the-art CNN classifiers to detect the COVID19 infection. However, the viral could of many different types; therefore, we only regard for COVID19 while the other viral infection types are treated as non-COVID19 in the radiographs of various viral infections. The classification task is challenging due to the limited number of scans available for COVID19 and the minute variations in the viral infections. We aim to employ current state-of-the-art CNN architectures, compare their results, and determine whether deep learning algorithms can handle the crisis appropriately.

Network

The network architectures and details are provided in our paper

Test

Quick start

  1. Download the trained models of our paper.

    The models and images for CT-COVID can be downloaded from Google Drive. The total size for zip is 4.96GB.

    The models and images for Xray-COVID can be downloaded from Google Drive. The total size for zip is 5.70GB. Please cite the respective datasets.Please cite the respective datasets.

  2. Cd to '/COVIDCT/CNN_methods' or '/COVIDXRAY/CNN_methods', then Cd in to the director of the method you want to test and run the following script.

    You can use the following script to test the algorithms

    bash predict_scripts.sh

Results

All the results for attention on the CT images can be downloaded from GoogleDrive from here, the images of attention for Xray images from here. The size of the results for CT-COVID19-Attention is about 400KB while for XRAY-COVID19-Attention is about 2.67MB

Visual Results

Representative images of the COVIDCT dataset employed for training and evaluation of algorithms

Following are the samples from the COVIDx dataset. The upper row shows the COVID19 infected examples, while the lower row presents the infection-free images

Next, we present the models on infected and non-infected radiographs. In Figure below, we present the CT images with feature attention where the red color indicates the region where the models have focused. The first three rows contain COVID19 infections, while the remaining two rows in Figure are infection-free.

Next Figure shows four different COVID19 infection radiographs from four different orientations. ResNet, DenseNet, and GoogleNet presented in the second, third, and fourth columns focused on most of the chest radiographs while the remaining models concentrated on particular regions of the chest. It is challenging for the models to pinpoint exactly the artifacts caused by COVID19, as is obvious from the feature attention mechanism.

Quantitative Results

Five quantitative measures for state-of-the-art deep learning networks on COVIDCT in the following table. The variants of the same algorithm are differentiated via the number at the end of the method’s name.

Quantitative results for state-of-the-art deep learning algorithms on COVIDx in the following table. The numbers at the end of the method name indicate variants of the same algorithms.

For more information, please refer to our paper

Citation

If you find the code helpful in your resarch or work, please cite the following papers.

@article{Anwar2020COVID19Detect,
       title={COVID19 detection from Radiographs: Is Deep Learning able to handle the crisis?},
       author={Saqib, Muhammad and Anwar, Saeed and Anwar, Abbas and Petersson, Lars and Sharma, Nabin and Blumenstein, Michael},  
       journal={Preprints-2020060189},
       year={2020},
}

About

COVID19 detection from Radiographs: Is Deep Learning able to handle the crisis ?

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published