Back
MRzero: Automated invention of MRI sequences using supervised learning
{We propose a framework \textemdash MRzero \textemdash that allows automatic invention of MR sequences. At the core of the framework is a differentiable forward process allowing to simulate image measurement and reconstruction. The sequence parameters are variables of optimization. As a cost function we use mean squared error distance to a certain given target contrast of interest. To avoid overfitting we propose a method that generates synthetic data that is used for training. In the experiments, we demonstrate the ability of the method to learn RF flip angles and spatial encoding from scratch given a target obtained with GRE sequence.}
@misc{item_3319823, title = {{MRzero: Automated invention of MRI sequences using supervised learning}}, booktitle = {{2021 ISMRM \& SMRT Annual Meeting \& Exhibition (ISMRM 2021)}}, abstract = {{We propose a framework \textemdash MRzero \textemdash that allows automatic invention of MR sequences. At the core of the framework is a differentiable forward process allowing to simulate image measurement and reconstruction. The sequence parameters are variables of optimization. As a cost function we use mean squared error distance to a certain given target contrast of interest. To avoid overfitting we propose a method that generates synthetic data that is used for training. In the experiments, we demonstrate the ability of the method to learn RF flip angles and spatial encoding from scratch given a target obtained with GRE sequence.}}, year = {2021}, slug = {item_3319823}, author = {Loktyushin, A and Herz, K and Dang, N and Glang, F and Deshmane, A and Weinm\"uller, F and Doerfler, A and Sch\"olkopf, B and Scheffler, K and Zaiss, M} }