Clinically significant Prostate Cancer Classification using Fusion

Reference Implementation of paper "Clinically Significant Prostate Cancer Detection using Multiparametric MRI: A Simple and Interpretable Deep Learning Method" of Heejong Kim, Himanshu Nagar, Daniel Margolis*, Mert Sabuncu* (*Senior author)

Dependencies

conda env create -f environment.yml
conda activate cspca-classification-fusion

PROSTATEx challenge Dataset Preparation

Please download the following files and change the file location variables to run the preprocessing script.

• Download the image SimpleITK transform files (Link)
• Download the PROSTATEx challenge dataset (Link)

python preprocessing.py

Training

We included all configuration files for

• One channel input (T2, ADC, DWIb800, Ktrans)
• Three channel input (T2-ADC-DWIb800, T2-ADC-Ktrans)

# For example, 
python train.py --config ./config/adc/augment-all-cnn3d8conv4mp3fcf4-1channel-batch64-adam-lre-0001-bcelogit-auc-nobfc.yaml

Testing

The testing script outputs accuracy metrics, such as AUC with confidence interval and saliency maps. Other than the analysis result of testset with labels, you can also get the PROSTATEx challenge testset submission results. Please note that challenge submission is not available as of Apr 30th. ([Challenge board] (https://prostatex.grand-challenge.org/))

Citation

If you use this code, please consider citing our work:

@article{kim2022pulse,
  title={Pulse Sequence Dependence of a Simple and Interpretable Deep Learning Method for Detection of Clinically Significant Prostate Cancer Using Multiparametric MRI},
  author={Kim, Heejong and Margolis, Daniel JA and Nagar, Himanshu and Sabuncu, Mert R},
  journal={Academic Radiology},
  year={2022},
  publisher={Elsevier}
}