Robust Compressed Sensing MR Imaging with Deep Generative PriorsDownload PDF

Published: 19 Oct 2021, Last Modified: 05 May 2023NeurIPS 2021 Deep Inverse Workshop PosterReaders: Everyone
Keywords: Generative priors, compressed sensing, MRI, Langevin dynamics, Posterior sampling, CSGM
TL;DR: Posterior sampling using score-based models is competitive with SOTA methods on fastMRI, and is robust to sampling and anatomy shift
Abstract: The CSGM framework (Bora-Jalal-Price-Dimakis'17) has shown that deep generative priors can be powerful tools for solving inverse problems. However, to date this framework has been empirically successful only on certain datasets (for example, human faces and MNIST digits), and it is known to perform poorly on out-of-distribution samples. In this paper, we present the first successful application of the CSGM framework on clinical MRI data. We train a generative prior on brain scans from the fastMRI dataset, and show that posterior sampling via Langevin dynamics achieves high quality reconstructions. Furthermore, our experiments and theory show that posterior sampling is robust to changes in the ground-truth distribution and measurement process.
Conference Poster: pdf
1 Reply