Skip to content

J-3TO/Sparse-View-Cranial-CT-Reconstruction

Repository files navigation

Improving Hemorrhage Detection in Sparse-view CTs via Deep Learning

Code to the paper "Improving Automated Hemorrhage Detection in Sparse-view CT via U-Net-based Artifact Reduction" https://pubs.rsna.org/doi/10.1148/ryai.230275

How to use this code:

  1. Download the stage_2_train folder from https://www.kaggle.com/competitions/rsna-intracranial-hemorrhage-detection/data
  2. Unzip and place in ./Data/
  3. Create dataset with the construct_dataset.ipynb
  4. Train U-Net with the train_U_Net.ipynb
  5. Train EfficientNet with train_EfficientNet.ipynb
  6. Calculate ROC curves with eval.ipynb
  7. Calculate SSIM/PSNR values with Calculate_SSIM_PSNR.ipynb

Main Dependencies:

The code was used on following libraries:

  • tensorflow==2.4.0
  • astra==2.1.0
  • pydicom==2.3.0
  • pandas==1.4.2
  • sklearn==1.1.3
  • skimage==0.19.3