smallmax00 / BAGCN-Covid19

MedIA 2022: code for 'Bilateral Adaptive Graph Convolutional Network on CT based COVID-19 Diagnosis with Uncertainty-Aware Consensus-Assisted Multiple Instance Learning Medical Image Analysis'

BAGCN-Covid19

Access to the data and code: https://doi.org/10.5281/zenodo.6361963

Dataset:

--We provide the annotated 7768 masks of C-19 CT slices, also we have provided the indices of the training&val, test datasets for two evaluations of learning ability (Cell) and generalization ability (cell+mosmed+covid-ct). The masks are in the file of masks.zip and the data indices are in the file of code_data_pre-trained-models.zip.

Implementation code:

Prepare your data, and split them as the index stated in ./data_index

UC-MIL

train:

run UC-MIL-train.py

BA-GCN

train:

prepare data for BA-GCN using the pre-trained UC-MIL model, select the top 16 slicesf ro each CT scan.

run BA-GCN-train.py

Pre-trained model.

We shared the pre-trained models of UC-MIL and BA-GCN in the folder ./pre-trained_model, for two evaluation settings called './Cell' and 'cell_mosmed_covid_ct'

Citation

If you find our work useful or use our dataset, please cite:

@article{meng2022bilateral,
  title={Bilateral adaptive graph convolutional network on CT based Covid-19 diagnosis with uncertainty-aware consensus-assisted multiple instance learning},
  author={Meng, Yanda and Bridge, Joshua and Addison, Cliff and Wang, Manhui and Merritt, Cristin and Franks, Stu and Mackey, Maria and Messenger, Steve and Sun, Renrong and Fitzmaurice, Thomas and others},
  journal={Medical Image Analysis},
  pages={102722},
  year={2022},
  publisher={Elsevier}
}