Meeting Banner
Abstract #3287

Alternating Joint Learning Approach for Variational Networks and Sampling Pattern in Parallel MRI

Marcelo Victor Wust Zibetti1, Florian Knoll2, and Ravinder Regatte1
1Radiology, NYU Langone Health, New York, NY, United States, 2Department of Artificial Intelligence in Biomedical Engineering, FAU Erlangen-Nuremberg, Erlangen, Germany

Synopsis

We propose a new alternating learning approach to jointly learn the sampling pattern (SP) and the parameters of a variational network (VN) for acquisition and reconstruction on 3D Cartesian parallel MRI problems. This approach is composed of alternating short training with BASS algorithm to learn the SP, and ADAM algorithm to learn the parameters of the VN, both with forced monotonicity. The results illustrate that this approach provides reduced error when compared to other joint learning approaches, and surpasses VN trained with recently developed fixed SPs.

This abstract and the presentation materials are available to members only; a login is required.

Join Here