DeepakRaya / CS_MRI_infoGAN_ADMM

Masters Thesis project

CS_MRI_infoGAN_ADMM

Load_data.py:

For converting and loading the data from neuroimaging format ".nii.gz" to pickel format, normalize the images in the range -1 to 1.

CS_MRI_ADMM_pytorch.ipynb:

Notebook contains the info-GAN model, Projector Model and their training routines.

neural_network_models.py:

The saved info-GAN generator and projector network are loaded in this module and composition of generator and projector network is performed, which is used as a denoiser in the second step of ADMM algorithm.

ADMM.py:

The Plug and Play ADMM solver algorithm program.

undersampling2.py:

Performs the undersampling in k-space of MRI image input and returns a Zero filled reconstruction.

stacked_optim.py:

Performs the first least squares like step of ADMM in a stacked column wise fashion.

SSIM.py:

For calculating the SSIM score between original and reconstructed images, using 11x11 gaussian window.

Experiment.ipynb:

Reconstrction experiments and resutls compilation notebook.