CAREER: Ultrahigh-Resolution Magnetic Resonance Spectroscopic Imaging for Label-Free Molecular Imaging of the Brain

了解大脑是如何工作的，包括如何有效地治疗大脑疾病，是最令人兴奋的科学前沿之一。神经成像通过允许以前所未有的细节对大脑解剖和活动进行无创测绘，显著地推进了这一前沿。然而，需要新的工具来可视化活大脑中的分子水平信息。目前的方法受到几个基本挑战的限制，包括需要放射性示踪剂或造影剂，有限的分子识别，噪声信号，成像时间长，空间分辨率差。这个CAREER项目的总体目标是开发新一代成像技术来解决这些挑战，并利用磁共振光谱成像（MRSI）以前所未有的空间和时间分辨率实现无标签的大脑分子成像。该研究计划的成功将显著推进分子神经成像领域，并通过同时绘制大脑中许多生理上重要的分子来实现新的功能。这些能力可能会彻底改变神经疾病和精神障碍的诊断和管理。教育活动将结合影像科学、机器学习和计算科学与工程交叉领域的研究与课程创新。活动包括开发本科生研究经验（REU）计划，为工程领域代表性不足的少数民族创造独特的生物医学成像培训和专业发展机会，以及以神经成像为中心的新的高中生研究实习计划。研究者的长期研究职业目标是开发新一代成像技术，以前所未有的时空分辨率实现大脑分子图谱的无标签映射，并探索这些技术在分子水平上研究大脑功能和疾病的潜力。为了实现这一目标，基于研究者的专业知识和之前在快速MRI和MRSI方面的贡献，这个CAREER项目的目标是开发一个创新的成像框架来建模、获取和处理MRSI数据，并实现新的基于MRSI的分子成像能力。核磁共振成像是一种潜在的强大的成像方式，允许在不需要放射性示踪剂和造影剂的情况下同时绘制人体许多生理上重要的分子。然而，由于核磁共振成像的信噪比低、速度慢、空间分辨率差、易受系统不完善的影响，迄今为止，核磁共振成像的发展仍处于起步阶段。该研究计划有五个目标：(1)通过整合生物先验、基于物理的建模和机器学习，为高维磁共振成像信号发现准确、高效的低维模型，以降低成像问题的维度，并在速度、分辨率和信噪比方面实现更好的权衡；(2)发展非常规的超快数据采集和数据处理策略，利用降维实现功能MRI分辨率水平下的全脑快速MRSI；(3)开发新的数学公式和高效算法，与新模型和获取协同工作，以实现最佳的空间光谱处理；(4)整合新的建模、获取和处理方法，实现对代谢物和神经递质及其生物物理特性（如弛豫和扩散参数）的全脑测绘，用于大脑的分子特异性微结构成像；(5)在拟议的MRSI框架中引入新的维度，以绘制分子依赖的生物物理特性的3D图谱。这些协同发展将使核磁共振成像从缓慢、低分辨率的模式转变为强大、高分辨率的体内分子神经成像工具，并为研究大脑生物化学、微观结构及其与功能和疾病过程的联系开辟了巨大的机会。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Understanding how the brain works, including how to effectively treat brain disorders, is one of the most exciting scientific frontiers. Neuroimaging has significantly advanced this frontier by allowing noninvasive mapping of the brain’s anatomy and activity in unprecedented detail. However, new tools are needed to visualize molecular-level information in a living brain. Current methods are limited by several fundamental challenges, including the need of radioactive tracers or contrast agents, limited molecule recognition, noisy signals, long imaging time, and poor spatial resolution. The overall goal of this CAREER project is to develop a new generation of imaging technologies to address these challenges and enable label-free molecular imaging of the brain at unprecedented resolutions in space and time using magnetic resonance spectroscopic imaging (MRSI). Success of the research planned will significantly advance the field of molecular neuroimaging and enable new capabilities by simultaneously mapping many physiologically important molecules in the brain. These capabilities could revolutionize diagnosis and management of neurological diseases and mental disorders. The educational activities will integrate the research with curriculum innovation at the intersections of imaging science, machine learning and computational science and engineering. Activities include development of a Research Experiences for Undergraduates (REU) program to create unique training and professional development opportunities in biomedical imaging for underrepresented minorities in engineering and a new high-school student research internship program centered on neuroimaging.The investigator’s long-term research career goal is to develop a new generation of imaging technologies to enable label-free mapping of molecular profiles in the brain at unprecedented spatiotemporal resolutions and explore the potential of these technologies for studying brain functions and diseases at the molecular level. Towards this goal, building on the investigator’s expertise and previous contributions on fast MRI and MRSI, the goal of this CAREER project is to develop an innovative imaging framework to model, acquire and process MRSI data and enable new MRSI based molecular imaging capabilities. MRSI is a potentially powerful imaging modality that allows for simultaneous mapping of many physiologically important molecules in the human body without the need of radioactive tracers and contrast agents. However, to date, the development of MRSI still remains at its infancy because of its low signal-to-noise ratio (SNR), slow speed, poor spatial resolution, and susceptibility to system imperfection. The Research Plan is organized under five objectives: (1) Discovery of accurate and efficient low-dimensional models for high-dimensional MRSI signals by integrating biological priors, physics-based modeling and machine learning, to reduce the dimensionality of the imaging problem and enable better tradeoffs in speed, resolution and SNR; (2) Development of unconventional ultrafast data acquisition and data processing strategies that exploit the reduced dimensionality to achieve fast MRSI of the whole brain at the resolution level of functional MRI; (3) Development of novel mathematical formulations and efficient algorithms that work synergistically with the new models and acquisitions for optimal spatiospectral processing; (4) Integration of the new modeling, acquisition and processing methods to enable whole-brain mapping of metabolites and neurotransmitters and their biophysical properties, such as relaxation and diffusion parameters, for molecule-specific microstructural imaging of the brain and (5) Introduction of new dimensions into the proposed MRSI framework to map molecule dependent biophysical properties in 3D. These synergistic developments will transform MRSI from a slow, low-resolution modality to a powerful, high-resolution in vivo molecular neuroimaging tool and open up tremendous opportunities in studying brain biochemistry, microstructure and their connections to functions and disease processes.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.