SpiNet

This repository contains the official implementation for SpiNet: A Deep Neural Network for Schatten p-norm Regularized Medical Image Reconstruction

This paper is published in Medical Physics journal. Please cite the paper if you intent on using the code.

Rastogi A, Yalavarthy PK. SpiNet: A deep neural network for Schatten p-norm regularized medical image reconstruction. Med Phys. 2021 May;48(5):2214-2229. doi: 10.1002/mp.14744. Epub 2021 Mar 22. PMID: 33525049.

Link for paper: https://pubmed.ncbi.nlm.nih.gov/33525049/

This code solves the following optimization problem:

J(x) = argmin_x ||Ax-b||_2^2 + ||x-Dw(x)||^p_p 

A can be any measurement operator. Here we consider parallel imaging problem in MRI where the A operator consists of undersampling mask, FFT, and coil sensitivity maps.

Dw(x): it represents the denoiser using a residual learning CNN.

Using Majorization-Minimization algorithm the optimization problem as be modified as:

 G(x|x*) = argmin_x ||Ax-b||_2^2 + p/2||W(x-Dw(x))||^p_p 
 
 W = diag((x*-Dw(x))^{p/2-1})

x*: is the point at which J(x*) = G(x*|x*)

Architecture

The architecture of our network is shown below

Training Data

The training data for SpiNet consists of brain scans of four patients and 360 images in total. The testing data consists of 164 images from one patients.

Hemant et al. [1]. have released the parallel imaging dataset used in their MoDL paper [1]. You can download the full dataset from the below link:

Download Link for file "dataset.hdf5" : https://drive.google.com/file/d/1qp-l9kJbRfQU1W5wCjOQZi7I3T6jwA37/view?usp=sharing

Files

The repository consists of the following files :-

1. R16.mat :- This file consists of the radial golden angle undersampling mask for 16 X undersampling rate.

2. MODL :- This folder consists of the saved model, training, testing, model generation and support funtion files for MODL. The files present in the folder are as follows :-

   • trn_old_R16.py :- This is the main training file for MODL
   • tst_old_R16.py :- This is the main testing file for MODL
   • model_old.py :- Model generating file for MODL
   • supportFunctions_old_R16.py :- Support function file for MODL used for reading and preprocessing training and testing data, generating PSNR values, loading saved model etc.

3. SpiNet :- This folder consists of the saved model, training, testing, model generation and support funtion files for SpiNet. The files present in the folder are as follows :-

   • trn_R16_8.py :- This is the main training file for SpiNet
   • tst_R16_8.py :- This is the main testing file for SpiNet
   • model_p_R6_8.py :- Model generating file for SpiNet
   • supportFunctions_R16.py :- Support function file for SpiNet used for reading and preprocessing training and testing data, generating PSNR values, loading saved model etc.

Note that dataset.hdf5 and R16.mat should be in MODL/SpiNet folder to while executing the training or testing code.

Contact The code is provided to support reproducible research. If the code is giving syntax error in your particular python configuration or some files are missing then you may open an issue or directly email me at adityar[at]iisc[dot]ac[dot]in

References

[1]. 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