Synthesizing raw multicoil MRI data with a deep generative framework.
Authors
Affiliations (3)
Affiliations (3)
- UC Berkeley-UCSF Graduate Program in Bioengineering, Berkeley, California, United States.
- University of California, Department of Radiology and Biomedical Imaging, San Francisco, California, United States.
- University of California, Department of Electrical Engineering and Computer Sciences, Berkeley, California, United States.
Abstract
Deep learning-based magnetic resonance imaging (MRI) reconstruction can enable faster imaging and improved image quality; however, clinical adoption is limited by the lack of diverse raw multicoil training datasets. In routine clinical workflows, multicoil <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>k</mi></mrow> </math> -space data needed for training are often discarded, and only magnitude images are retained. In addition, phase is not deterministically recoverable from magnitude, and storing full <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>k</mi></mrow> </math> -space is impractical. We address these limitations by (1) using a deep generative model to produce multiple plausible synthetic phase images conditioned on a single magnitude image and (2) introducing a radiofrequency (RF) coil library to synthesize multicoil k-space from the generated complex-valued images. The synthetic phase images resemble ground-truth phase as assessed by Fréchet inception distance, and the generated multicoil <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>k</mi></mrow> </math> -space effectively trains unrolled MRI reconstruction networks. These results indicate that generative phase synthesis coupled with an RF coil library provides a practical alternative to real multicoil <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>k</mi></mrow> </math> -space for training reconstruction models.