Image Reconstruction for MRI using Deep CNN Priors Trained without Groundtruth

TL;DR

This paper introduces a novel MRI reconstruction method that uses deep CNN priors trained without groundtruth, effectively removing artifacts from undersampled data and producing high-quality 4D images.

Contribution

The method uniquely trains CNN priors without artifact-free groundtruth, enabling effective MRI reconstruction from severely undersampled data.

Findings

High-quality 4D MRI images from severely undersampled data

Competitive performance against TGV and UNet3D methods

Effective removal of undersampling artifacts without groundtruth

Abstract

We propose a new plug-and-play priors (PnP) based MR image reconstruction method that systematically enforces data consistency while also exploiting deep-learning priors. Our prior is specified through a convolutional neural network (CNN) trained without any artifact-free ground truth to remove undersampling artifacts from MR images. The results on reconstructing free-breathing MRI data into ten respiratory phases show that the method can form high-quality 4D images from severely undersampled measurements corresponding to acquisitions of about 1 and 2 minutes in length. The results also highlight the competitive performance of the method compared to several popular alternatives, including the TGV regularization and traditional UNet3D.