Skip to content

Sparse-View CT Reconstruction with Deep Convolutional Neural Network (Project): Created a deep convolutional neural network (FBPConvNet based on U-Net) to reconstruct CT images from sparse-view data, achieving significant gains in image quality (PSNR, SSIM) over traditional methods through systematic fine-tuning. PDF report is in the report folder.

License

Notifications You must be signed in to change notification settings

JD9898/CT_reconstruction_with_FBPConvNet

Repository files navigation

MPhil DIS Thesis Project

Description + motivation

This project is for the MPhil DIS thesis.

Data

The data has been omitted in this repo for storage consideration but can be accessed at: LoDoPaB dataset: https://drive.google.com/drive/folders/1j2ODL6vimbv9KmXNvHZFvrJ4xGV7vDyJ?usp=drive_link Ellipsoidal dataset: https://drive.google.com/drive/folders/1VV9kMhHCSrNpGk2Sdt4wnjiIex_PXdQH?usp=drive_link Please make sure to install these datasets in the root directory mw918 before attempting to run the code. If the links fail to work, please contact mw918@cam.ac.uk

Installation and Usage of HPC to run the program

  1. First log in to HPC,
ssh <username>@login.hpc.cam.ac.uk
  1. Then move the project mw918 onto HPC,
  2. Navigate to the project directory: cd mw918
  3. Set up environment
conda env create --file=environment.yml
conda activate project
  1. Change the submit.job file accordingly. Then on HPC terminal:
sbatch submit.job

About

Sparse-View CT Reconstruction with Deep Convolutional Neural Network (Project): Created a deep convolutional neural network (FBPConvNet based on U-Net) to reconstruct CT images from sparse-view data, achieving significant gains in image quality (PSNR, SSIM) over traditional methods through systematic fine-tuning. PDF report is in the report folder.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published