Skip to content

Latest commit

 

History

History
35 lines (26 loc) · 3.19 KB

README.md

File metadata and controls

35 lines (26 loc) · 3.19 KB
Raw

This is a repository for Chest X-Ray COVID-19 detection based on a COvidNet and different preprocessing steps. The network implementation is based on the COVIDNet proposed in L. Wang, Z. Q. Lin, and A. Wong, “Covid-net: A tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images,” Scientific Reports, vol. 10, no. 19549, 2020, following the PyTorch implementation available at GitHub

The respository includes also impelementation of a network based on ResNet50, as proposed in: Li, L., Qin, L., Xu, Z., Yin, Y., Wang, X., Kong, B., Bai, J., Lu, Y., Fang, Z., Song, Q. and Cao, K., 2020. Using artificial intelligence to detect COVID-19 and community-acquired pneumonia based on pulmonary CT: evaluation of the diagnostic accuracy. Radiology, 296(2). url

The models available are:

  • Original CovidNet
  • Modified CovidNet with dropout and Grad_CAM functionalities
  • CovidNet with Deep Explainer functionalities
  • Covid_Resnet50
  • DenseNet

The preprocessing steps evaluated include cropping and segmentation of the lungs based on the 2D Lung segmentation available at GitHub. This network is also included in the segmenter directory of this respository. The description of all the experiments and results can be consulted on:

J.D. Arias-Londoño, J.A. Gómez-García, L. Moro-Velázquez, J.I. Godino-Llorente. Artificial Intelligence applied to chest X-Ray images for the automatic detection of COVID-19. A thoughtful evaluation approach. IEEE Access, vol 8. 2020. Open access

If you find our work useful, can cite our paper using:

@article{arias2020,
  title={Artificial Intelligence applied to chest X-Ray images for the automatic detection of COVID-19. A thoughtful evaluation approach},
  author={Arias-Londo{\~n}o, Julian D and Gomez-Garcia, Jorge A and Moro-Velazquez, Laureano and Godino-Llorente, Juan I},
  journal={IEEE Access},
  volume = {8},
  year={2020}
}

To execute the code take into account the following steps:

  • Build the train.txt and test.txt files according to the structure proposed in the original implementation, which include the name of every file with its corresponding label ['pneumonia', 'normal', 'COVID-19'].
  • The path to those files must be set in dataset.py file
  • To reproduce experiment 1 of the paper run main.py with the original images
  • To reproduce experiments 2 and 3, you can use LungSegmentation notebook to apply the Segmenter CNN to the dataset and get a segmentation mask for every single image in the dataset.
  • In the MatLab_preprocessing directory there is a script that takes the masks obtained in the previous step and apply them to the original images in order to get the crooped and crooped-segmented datasets; this preprocessed images can be used to reproduce expriments 2 and 3 of the paper respectively. For doing that, just change the path to the data directory during the call to main.py, and point it to the directories containing the preprocessed images.