VORTEX: Physics-Driven Data Augmentations for Consistency Training for Robust Accelerated MRI Reconstruction.
Year: 2021
Type: article
Abstract: Deep neural networks have enabled improved image quality and fast inference \ntimes for various inverse problems, including accelerated magnetic resonance \nimaging (MRI) reconstruction. However, such models require large amounts of \nfully-sampled ground truth data, which are difficult to curate and are \nsensitive to distribution drifts. In this work, we propose applying \nphysics-driven data augmentations for consistency training that leverage our \ndomain knowledge of the forward MRI data acquisition process and MRI physics \nfor improved data efficiency and robustness to clinically-relevant distribution \ndrifts. Our approach, termed VORTEX (1) demonstrates strong improvements over \nsupervised baselines with and without augmentation in robustness to \nsignal-to-noise ratio change and motion corruption in data-limited regimes; (2) \nconsiderably outperforms state-of-the-art data augmentation techniques that are \npurely image-based on both in-distribution and out-of-distribution data; and \n(3) enables composing heterogeneous image-based and physics-driven \naugmentations.
Institution: Stanford University
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W3211257574