CellSegDA

This repo holds code for Adversarial Domain Adaptation for Cell Segmentation

Usage

1. Environment

Run following commands to prepare environment with all dependencies.

conda env create -f environment.yml
conda activate cellseg-da

2. Dataset

Please send an email to mohammadminhazu.haq AT mavs.uta.edu to request the datasets.

3. Training

CellSegUDA

Run following script to train CellSegUDA model with KIRC as source dataset and TNBC as target dataset.

python train_cellseg_uda.py --source_dataset kirc --target_dataset tnbc

CellSegSSDA

Run following script to train CellSegSSDA model with KIRC as source dataset, TNBC as target dataset, and 25% labels of target dataset.

python train_cellseg_ssda.py --source_dataset kirc --target_dataset tnbc --target_label_percentage 25

4. Prediction and Evaluation

Run following script to predict the segmentation masks on kirc-test images, and then evaluate the predictions.

python predict_and_evaluate.py --test_dataset kirc --model_path path_to_best_model

Citations

@inproceedings{haq2020adversarial,
  title={Adversarial domain adaptation for cell segmentation},
  author={Haq, Mohammad Minhazul and Huang, Junzhou},
  booktitle={Medical Imaging with Deep Learning},
  pages={277--287},
  year={2020},
  organization={PMLR}
}