On Instabilities of Conventional Multi-Coil MRI Reconstruction to Small Adverserial Perturbations

TL;DR

This paper investigates the vulnerability of multi-coil MRI reconstruction methods to small adversarial perturbations, revealing significant instabilities that raise concerns for clinical deployment of deep learning-based MRI techniques.

Contribution

It is the first to analyze adversarial instabilities in multi-coil MRI reconstruction, extending prior single-coil studies and highlighting practical risks.

Findings

Multi-coil MRI reconstructions are highly susceptible to small adversarial attacks.

Parallel imaging and compressed sensing methods exhibit notable instabilities.

Results emphasize the need for robustness in deep learning MRI methods.

Abstract

Although deep learning (DL) has received much attention in accelerated MRI, recent studies suggest small perturbations may lead to instabilities in DL-based reconstructions, leading to concern for their clinical application. However, these works focus on single-coil acquisitions, which is not practical. We investigate instabilities caused by small adversarial attacks for multi-coil acquisitions. Our results suggest that, parallel imaging and multi-coil CS exhibit considerable instabilities against small adversarial perturbations.