AI-Driven Design of RF Pulses for Enhancing Nuclear Magnetic Resonance Spectroscopy and Imaging

在化学系化学测量和成像(CMI)计划的支持下，明尼苏达大学的Gianluigi Veglia教授和他的团队正在开发新的射频(RF)辐照技术，以增强重要的化学结构探针的性能，包括核磁共振(NMR)光谱和磁共振成像(MRI)。具体地说，Veglia团队正在利用功能强大的软件来调用人工智能来协助设计过程。由此产生的方法有可能获得更高质量的数据，用于生物聚合物的化学和结构分析，因此，有望改善磁共振成像的图像质量和采集时间。开发的方法将免费提供给科学界。从事这项研究的本科生和研究生将接受磁共振基本和高级概念的严格培训，并介绍如何使用人工智能进行优化过程。Veglia团队最近开发了新的软件(Genetics-AI)，该软件可以生成高度补偿的RF脉冲形状，并具有前所未有的保真度，适用于核磁共振和核磁共振。新的射频形状本质上是宽带的，具有可调的自旋操作保真度。新的射频脉冲已经使用经典的自旋纠缠问题进行了基准测试，获得了99.999%的保真度。在液体和固态核磁共振实验中的应用导致了多维谱的更高的信噪比水平。该团队目前正在使用进化算法和随后的人工智能例程训练来开发同时具有相位和幅度调制的射频脉冲的能力，以优化新射频形状的功率和相位。最终目标是获得高保真脉冲和脉冲序列，这些脉冲和脉冲序列将在商业核磁共振光谱仪或在高和超高磁场下运行的核磁共振扫描仪上实施。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

With support from the Chemical Measurement and Imaging (CMI) Program in the Division of Chemistry, Professor Gianluigi Veglia and his group at the University of Minnesota are developing new radiofrequency (RF) irradiation techniques to enhance the performance of important probes of chemical structure, including nuclear magnetic resonance (NMR) spectroscopy and magnetic resonance imaging (MRI). Specifically, the Veglia group is utilizing powerful software that invokes artificial intelligence to assist the design process. The resulting methods have the potential to enable the acquisition of higher quality data for chemical and structural analysis of biopolymers and, as a result, are expected to improve image quality and acquisition times in MRI. The methods developed will be made freely available to the scientific community. Undergraduate and graduate students engaged in the research will receive rigorous training in basic and advanced concepts in magnetic resonance and an introduction to the use of artificial intelligence for optimization processes. The Veglia group has recently developed new software (GENETICS-AI) that generates highly compensated RF pulse shapes with unprecedented levels of fidelity for applications to NMR and MRI. The new RF shapes are inherently broadband, with tunable fidelity of spin operation. The new RF pulses have been benchmarked using a classical spin entanglement problem, obtaining a fidelity level of 99.999%. The application to liquid and solid-state NMR experiments resulted in higher signal-to-noise levels for multidimensional spectra. The team is now developing capabilities for RF pulses with both phase and amplitude modulations using an evolutionary algorithm and subsequent training of an artificial intelligence routine to optimize the power and phase of new RF shapes. The ultimate goal is to obtain high-fidelity pulses and pulse sequences that will be implemented on commercial NMR spectrometers or MRI scanners operating at high and ultra-high magnetic fields.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.