Short Communication: Localized Adversarial Artifacts for Compressed Sensing MRI

As interest in deep neural networks (DNNs) for image reconstruction tasks grows, their reliability has been called into question [V. Antun, F. Renna, C. Poon, B. Adcock, and A. C. Hansen, Proc. Natl. Acad. Sci. USA, 117 (2020), pp. 30088-30095; N. M. Gottschling, V. Antun, B. Adcock, and A. C. Hansen, The Troublesome Kernel: Why Deep Learning for Inverse Problems Is Typically Unstable, preprint, arXiv:2001.01258, 2020]. However, recent work has shown that, compared to total variation (TV) minimization, when appropriately regularized, DNNs show similar robustness to adversarial noise in terms of l(2)-reconstruction error [M. Genzel, J. Macdonald, and M. Marz, IEEE Trans. Pattern Anal., 45 (2022), pp. 1119-1134]. We consider a different notion of robustness, using the l(infinity)-norm, and argue that localized reconstruction artifacts are a more relevant defect than the l(2)-error. We create adversarial perturbations to undersampled magnetic resonance imaging measurements (in the frequency domain) which induce severe localized artifacts in the TV-regularized reconstruction. Notably, the same attack method is not as effective against DNN-based reconstruction. Finally, we show that this phenomenon is inherent to reconstruction methods for which exact recovery can be guaranteed, as with compressed sensing reconstructions with l(1)- or TV-minimization.