Accelerated motion correction with deep generative diffusion models

Abstract

We propose a flexible framework for retrospective motion correction of accelerated MRI based on deep generative diffusion models, with potential application to other forward model corruptions.