Classification of disease severity using the COV19-CT-DS

COV19-CT-DB

Data folders and file counts

Summary:

Diagnosis	Subset	n	Count per severity class
noncovid	train		N/A
	val		N/A
	test		N/A
covid	train	460	132, 123, 166, 39
	val	101	31, 20, 45, 5
	test	231	N/A

Reproducing results

1. Installation

pip install -r requirements.txt 

2a. Run interactively in ipython or python

$ ipython


In [1]: BS=128; IMGSET='all2'; MID='wideres101'; exec( open('icassp_sep.py').read() )

2b. Run non-interactively

$ python icassp_sep.py 1 0 16   

# MID=1
#
# FT=0 # FT=0 : Fine tune the last layer only
#      # FT=1 : Fine tune all layers
#
# BS=16 

Thank you for your exploration of this repo and consideration to cite this work:

Tang LY. Severity classification of ground-glass opacity via 2-D convolutional neural network and lung CT scans: a 3-day exploration. arXiv preprint arXiv:2303.16904. 2023 Mar 23.

@article{tang2023severity,
  title={Severity classification of ground-glass opacity via 2-D convolutional neural network and lung CT scans: a 3-day exploration},
  author={Tang, Lisa YW},
  journal={arXiv preprint arXiv:2303.16904},
  year={2023}
}

Other resources

Datasets

• https://www.eibir.org/covid-19-imaging-datasets/
• https://www.kaggle.com/code/utkarshsaxenadn/ct-scans-3d-data-3d-data-processing-3d-cnn#3D-Scans-Data-Loading

Lung mask segmentation code

https://github.com/pzaffino/COVID19-intensity-labeling/blob/main/lungs_processing.py

• uses Simple ITK