High signal‐to‐noise ratio (SNR) of the NMR signal has always been a key target that drives massive research effort in many fields. Among several parameters, a high filling factor of the MR coil has proven to boost the SNR. In case of small‐volume samples, a high filling factor and thus a high SNR can be achieved through miniaturizing the MR coil. However, under certain circumstances, this can be impractical. In this paper, we present an extensive theoretical and experimental investigation of the inductively coupled LC resonator and the magnetic Lenz lens as two candidate approaches that can enhance the SNR in such circumstances. The results demonstrate that the narrow‐band LC resonator is superior in terms of SNR, while the non‐tuned nature of the Lenz lens makes it preferable in broadband applications.

Two major technical challenges facing parallel nuclear magnetic resonance (NMR) spectroscopy, at the onset, include the need to achieve exceptional $$B_0$$B 0 homogeneity, and good inter-detector radiofrequency signal decoupling, and have remained as technical obstacles that limit high throughput compound screening via NMR. In this contribution, we consider a compact detector system, consisting of two NMR ‘unit cell’ resonators that implement parallel $$B_0$$B 0 shimming with parallel radiofrequency detection, as a prototype NMR environment, pointing the way towards achieving accelerated NMR analysis. The utility of our approach is established by achieving local field correction within the bore of a 1.05T permanent magnet MRI. Our forerunner platform suppresses signal cross-coupling in the range of $$-21$$- 21 dB to $$-30$$- 30 dB, under a geometrically decoupled scheme, leading to a halving of the necessary inter-coil separation. In this permanent magnet environment, two decoupled parallel NMR detector sites simultaneously achieve narrow spectral linewidth, overcoming the spatial inhomogeneity of the magnet from 400 to 28 Hz.

We introduce a low-complexity, low-cost, yet sufficiently accurate automatic tune and match system for NMR and MRI applications. The ArduiTaM builds upon an Arduino Uno embedded system that drives a commercial frequency synthesiser chip to perform a frequency sweep around the Larmor frequency. The generated low-power signal is fed to the NMR coil, after which the reflected waves are detected using a directional coupler and amplified. The signal shape is then extracted by means of an envelope detector and passed on to the Arduino, which performs a dip search while continuously generating actuator control patterns to adjust the tune and match capacitors. The process stops once the signal dip reaches the Larmor frequency. The ArduiTaM works readily with any spectrometer frequency in the range from 1 to 23 T. The speed of the ArduiTaM is mainly limited by the clock of the Arduino and the capacitor actuation mechanism. The Arduino can easily be replaced by a higher-speed microcontroller, and varactors can replace stepper-motor controlled variable capacitors. The ArduiTaM is made available in open source, and so is easily duplicated.

Shimming in the context of nuclear magnetic resonance aims to achieve a uniform magnetic field distribution, as perfect as possible, and is crucial for useful spectroscopy and imaging. Currently, shimming precedes most acquisition procedures in the laboratory, and this mostly semi-automatic procedure often needs to be repeated, which can be cumbersome and time-consuming. The paper investigates the feasibility of completely automating and accelerating the shimming procedure by applying deep learning (DL). We show that DL can relate measured spectral shape to shim current specifications and thus rapidly predict three shim currents simultaneously, given only four input spectra. Due to the lack of accessible data for developing shimming algorithms, we also introduce a database that served as our DL training set, and allows inference of changes to ¹H NMR signals depending on shim offsets. In situ experiments of deep regression with ensembles demonstrate a high success rate in spectral quality improvement for random shim distortions over different neural architectures and chemical substances. This paper presents a proof-of-concept that machine learning can simplify and accelerate the shimming problem, either as a stand-alone method, or in combination with traditional shimming methods. Our database and code are publicly available.