Learning MRI Artifact Removal With Unpaired Data

TL;DR

This paper introduces a neural network approach for MRI artifact removal that learns from unpaired data, enabling effective artifact correction without the need for paired training datasets.

Contribution

It presents a novel unpaired learning method for MRI artifact removal, eliminating the need for paired datasets and improving practical applicability.

Findings

Effective artifact removal demonstrated across different MRI contrasts.

Retains anatomical details while removing artifacts.

Works without paired artifact-free and corrupted images.

Abstract

Retrospective artifact correction (RAC) improves image quality post acquisition and enhances image usability. Recent machine learning driven techniques for RAC are predominantly based on supervised learning and therefore practical utility can be limited as data with paired artifact-free and artifact-corrupted images are typically insufficient or even non-existent. Here we show that unwanted image artifacts can be disentangled and removed from an image via an RAC neural network learned with unpaired data. This implies that our method does not require matching artifact-corrupted data to be either collected via acquisition or generated via simulation. Experimental results demonstrate that our method is remarkably effective in removing artifacts and retaining anatomical details in images with different contrasts.