Acquisitions with random shim values enhance AI-driven NMR shimming

Moritz Becker; Sören Lehmkuhl; Stefan Kesselheim; Jan G Korvink; Mazin Jouda

doi:10.1016/j.jmr.2022.107323

Acquisitions with random shim values enhance AI-driven NMR shimming

J Magn Reson. 2022 Dec:345:107323. doi: 10.1016/j.jmr.2022.107323. Epub 2022 Oct 30.

Authors

Moritz Becker¹, Sören Lehmkuhl¹, Stefan Kesselheim², Jan G Korvink³, Mazin Jouda⁴

Affiliations

¹ Karlsruhe Institute of Technology (KIT), Institute of Microstructure Technology, Karlsruhe 76131, Germany.
² Forschungszentrum Jülich (FZJ), Jülich Supercomputing Centre, Jülich 52428, Germany.
³ Karlsruhe Institute of Technology (KIT), Institute of Microstructure Technology, Karlsruhe 76131, Germany. Electronic address: jan.korvink@kit.edu.
⁴ Karlsruhe Institute of Technology (KIT), Institute of Microstructure Technology, Karlsruhe 76131, Germany. Electronic address: mazin.jouda@kit.edu.

PMID: 36375285
DOI: 10.1016/j.jmr.2022.107323

Abstract

Shimming is still an unavoidable, time-consuming and cumbersome burden that precedes NMR experiments, and aims to achieve a homogeneous magnetic field distribution, which is required for expressive spectroscopy measurements. This study presents multiple enhancements to AI-driven shimming. We achieve fast, quasi-iterative shimming on multiple shims simultaneously via a temporal history that combines spectra and past shim actions. Moreover, we enable efficient data collection by randomized dataset acquisition, allowing scalability to higher-order shims. Application at a low-field benchtop magnet reduces the linewidth in 87 of 100 random distortions from ∼ 4 Hz to below 1 Hz, within less than 10 NMR acquisitions. Compared to, and combined with, traditional methods, we significantly enhance both the speed and performance of shimming algorithms. In particular, AI-driven shimming needs roughly 1/3 acquisitions, and helps to avoid local minima in 96% of the cases. Our dataset and code is publicly available.

Keywords: AI-driven NMR Shimming; Automated Shimming; Deep Learning; Nuclear Magnetic Resonance.