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:

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

The code was used on following libraries:

Name		Name	Last commit message	Last commit date
Latest commit History 41 Commits
Data		Data
model_weights/EfficientNet		model_weights/EfficientNet
Calculate_SSIM_PSNR.ipynb		Calculate_SSIM_PSNR.ipynb
README.md		README.md
construct_dataset.ipynb		construct_dataset.ipynb
eval.ipynb		eval.ipynb
models.py		models.py
test_data.csv		test_data.csv
train_EfficientNet.ipynb		train_EfficientNet.ipynb
train_UNet.csv		train_UNet.csv
train_U_Net.ipynb		train_U_Net.ipynb
utils.py		utils.py
val_UNet.csv		val_UNet.csv

Provide feedback