nahia1 / tfg-cnn

Deep learning scatter correction for chest x-rays

tfg-cnn

PyTorch implementation of a convolutional neural network for scatter correction in chest x-rays.

Dependencies

• torch==1.11.0
• torchvision==0.12.0
• numpy==1.22.3
• opencv_python==4.5.5.64

Dataset

NIH Chest X-ray dataset: https://nihcc.app.box.com/v/ChestXray-NIHCC

Usage

Training example

python3 train.py --epochs 20 --patch-size 128 --learning-rate 1e-4 dataset/train

Testing example

python3 eval.py --patch-size 128 --patch-stride 64 trained.pth dataset/test

Examples

• Left: ground truth (no scatter)
• Middle: input (with scatter)
• Right: output/prediction (scatter corrected)