This is a self-implemented torch version for MoDL. For further information and original implementation, please refer to:
https://github.com/hkaggarwal/modl
I also separate the CGSENSE procedure in the CGSENSE.py
Different from the original code, I utilize first 50 slices in the test dataset for validation. Hope it doesn't bother. -- please check util.py
I tried to mimic the tensorflow version using the torch for brain dataset, but batch normalization did not work, it changed the scale of the data and deterioated. If you figure out, please feel free to contact me.
MoDL: Model Based Deep Learning Architecture for Inverse Problems by H.K. Aggarwal, M.P Mani, and Mathews Jacob in IEEE Transactions on Medical Imaging, 2018
Link: https://arxiv.org/abs/1712.02862
IEEE Xplore: https://ieeexplore.ieee.org/document/8434321/
Presentation: https://github.com/hkaggarwal/modl/blob/master/MoDL_ppt.pdf
From the original link:
Download Link : https://drive.google.com/file/d/1qp-l9kJbRfQU1W5wCjOQZi7I3T6jwA37/view?usp=sharing