/COV19D_4th

A mobile-friendly solution for COVID-19 diagnosis from CT images using Mobile ViT transformers

Primary LanguagePython

DataBase

COV19-CT Database was shared in the forth run of the competition and can be obtained from the workshop organizers at https://mlearn.lincoln.ac.uk/ai-mia-cov19d-competition/.

Method

This code can be deployed in either of two ways: without CT images processing (you may skip this step in the code) or with images processing:

Images Processing (Optional). Images were processed by deleting non-representative slices in each CT scan, and cropping the Region Of Interest (ROI) , i.e. the lung areas.

Vision Trnasformer for Slices Diagnosis. Vision Trnasformer-based methodology (xxs mobile ViT Transformer) was used to make diagnosis decisions at the slice level. Next, majority voting was used to make the final diagnostic decisions for each patient.

  • Please note: This is a binary classification task. To replicate the method on multi-class classification data, you need to modify the model's output to suit your task.
  • Please refer to the attached paper for more details on the methodology.
  • Kindly inform the organization owner if you wish to obtain the pretrained model in this study.

Dependencies

torch==1.10.1
torchvision==0.11.2
timm==0.6.12
pil==8.3.1

Citation

If you find the this method useful, please consider citing:

@misc{morani2023covid19,
title={COVID-19 Detection Using Swin Transformer Approach from Computed Tomography Images},
author={Kenan Morani},
year={2023},
eprint={2310.08165},
archivePrefix={arXiv},
primaryClass={eess.IV}
}