- The code is for denoising of MR images using deep learning referred to as Coil2Coil (C2C). C2C generated paired noise-corrupted images from phased-array coil data to train a deep neural network, and the paried images were modified to satisfy the conditions of Noise2Noise (N2N; Lehtinen, et al), enabling network training using N2N algorithm.
- last update : 2022. 08. 07
- will be updated
-
Python 3.7
-
pytorch=1.9.0
-
NVIDIA GPU
First, download the codes using the command below.
git clone https://github.com/SNU-LIST/Coil2Coil.git
Then, go inside 'Coil2Coil' file, and install the ssmd packages
Following command will be install a few dependent libraries such as Pytorch, Numpy, and etc...
pip install -e denoise
There might be 'denoise' library on your environment (pip list)
Run the eval.py with following codes
Arguments are provided below:
python eval.py -dir {directory to save logs}
-m {model weight path}
-d {dataset path for inference}
-noise {amount of noise (inf for just inference)}
Or if you just wanna inference one .mat file ...
python test_chan1.py -dir {directory to save logs}
-m {model weight path}
-d {dataset path for inference}
-noise {amount of noise (inf for just inference)}
Run the train.py with following codes
Arguments are provided below:
python train.py -dir {directory to save logs}
-a {algorithm, 'c2c' or 'n2c'}
-t {path for training dataset}
-v {path for validation dataset}
-noise {amount of noise (inf or gauss(float))}
We provide software for academic research purpose only and NOT for commercial or clinical use.
For commercial use of our software, contact us (snu.list.software@gmail.com) for licensing via Seoul National University.
Please email to “snu.list.software@gmail.com” with the following information.
Name:
Affiliation:
Software:
When sending an email, an academic e-mail address (e.g. .edu, .ac.) is required.
Juhyug Park, M.S-Ph.D candidate, Seoul National University