CytoSet is a deep-learning based method used for predicting clinical outcome from cytometry data.
- Python >= 3.6
- CUDA >= 10.1
pip install -r requirements.txt
AML dataset is from https://flowrepository.org/id/FR-FCM-ZZYA. HEUvsUE dataset is from https://flowrepository.org/id/FR-FCM-ZZZU. ICS dataset is from https://flowrepository.org/id/FR-FCM-ZZZV. NK_cell dataset is from the repository of CellCNN.
The pre-processed dataset for training the model can be downloaded from the google drive.
- Download pre-processed the datasets (see Datasets Section) and unpack them.
- In
scripts/train/train_[Dataset].sh
, setbin_file
to the path oftrain.py
andgpu
to the gpu id. - Start training:
bash train_[Dataset].sh
- We provide our pre-trained model on HVTN dataset and test dataset in
checkpoints
. - We also provide our model configuration for each dataset in
config/model
. - To run the testing, you can use the following command:
python test.py --model checkpoints/HVTN_model.pt --config config/model/ICS/config.json --test_pkl checkpoints/test_sample.pkl
The evaluation results are:
Accuracy | Area Under Curve |
---|---|
0.958 | 0.962 |
@inproceedings{
10.1145/3459930.3469529,
author = {Yi, Haidong and Stanley, Natalie},
title = {CytoSet: Predicting Clinical Outcomes via Set-Modeling of Cytometry Data},
year = {2021},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3459930.3469529}
}
If you have any questions, please feel free to contact Haidong Yi ([email protected]) or push an issue on Issues Dashboard.