Simultaneous Multinuclear Magnetic Resonance Fingerprinting for Data Fusion of Quantitative Structural and Metabolic Imaging

用于定量结构和代谢成像数据融合的同步多核磁共振指纹图谱

• 批准号: 9889957
• 负责人: Guillaume MADELIN
• 金额: $ 64.7万
• 依托单位: NEW YORK UNIVERSITY SCHOOL OF MEDICINE
• 依托单位国家: 美国
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-07-01 至 2022-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9889957
• 关键词: 3-Dimensional; Affect; Age; Algorithms; Anterior; Architecture; Back; Biochemistry; Blood Circulation; Brain; Brain Ischemia; Cerebrospinal Fluid; Chronic; Clinical; Computer software; Data; Development; Diffusion; Disease; Ensure; Exhibits; Extracellular Fluid; Fingerprint; Frequencies; Gender; Goals; Homeostasis; Human; Image; Imaging Techniques; Intracellular Fluid; Ions; MRI Scans; Magnetic Resonance; Maps; Measurable; Measurement; Metabolic; Metabolism; Methods; Minor; Modality; Modeling; Monitor; Morphology; Nature; Nerve Degeneration; Neurocognitive Deficit; Patients; Perfusion; Physiologic pulse; Play; Process; Property; Protocols documentation; Protons; Recurrence; Reproducibility; Resolution; Scanning; Signal Transduction; Sodium; Stroke; Structure; Techniques; Testing; Time; Tissue Model; Tissues; Training; Transient Ischemic Attack; Validation; base; clinical application; computerized data processing; data acquisition; data fusion; density; design; extracellular; flexibility; hemodynamics; in vivo; insight; learning strategy; metabolic abnormality assessment; metabolic imaging; models and simulation; molecular imaging; new technology; non-invasive imaging; novel; prototype; radio frequency; reconstruction; simulation; statistical learning

Project Summary In this project we want to develop a new non-invasive imaging technique that will provide multi-parametric metabolic maps of the living brain at an unprecedented resolution. The key to this new technology is the novel combination of three state-of-the-art imaging concepts: (A) new hardware that enables the simultaneous measurement of multinuclear magnetic resonance (MR) signals at different frequencies; (B) the flexibility and robustness of Plug-and-Play (PnP) MR Fingerprinting (MRF); and (C) a data fusion process driven by a cross-modality model based on statistical learning. For brevity, we will call this fused simultaneous multinuclear PnP-MRF technique MNF (Multi-Nuclear Fusion). The idea behind MNF is to rapidly capture two different kinds of quantitative information throughout the whole brain in one single scan: (1) structural information from proton (1H) MRF such as T1, T2, and proton density (PD) (tissue-scale morphology); and (2) metabolic information related to ion homeostasis from sodium (23Na) MRF, such as intracellular sodium concentration, and intracellular, extracellular and cerebrospinal fluid (CSF) volume fractions (cellular-scale function). Because PnP-MRF can quantify multiple tissue properties free of experimental bias, it enables us to employ statistical learning to discover a subject-specific cross-modality model that integrates all voxelwise inter-relationships between the multi-parametric 1H PnP-MRF (acquired at high resolution, 0.75-1 mm) and 23Na PnP-MRF (acquired at low resolution, 3-5 mm) maps. These subject-specific relations can subsequently be used to sharpen the 23Na metabolic maps to match the resolution of the 1H structural maps. The high-resolution 23Na maps will enable the assessment of metabolic processes in vivo and bridge the gap in resolution that has held back our ability to study metabolism in the living human brain, which is crucial for our understanding of the brain itself and the afflictions that affect it. This proof-of-concept implementation will be developed at 7 T, but it is expected to be adaptable to clinical 3 T MR scanners. The specific aims are: (1) Data acquisition, (1.a) multi-channel 1H/23Na RF array, (1.b) simultaneous multinuclear 3D MRF sequence; (2) Data processing, (2.a) PnP-MRF reconstruction for both 1H data (fingerprint matching to generate structural maps) and 23Na data (tissue 4-compartment model and simulation of spin 3/2 dynamics to generate metabolic maps), (2.b) cross-modality model using statistical learning, and data fusion algorithm to generate high-resolution metabolic maps; (3) Method validation, (3.a) accuracy and precision, (3.b) repeatability and reproducibility.(3) Exploratory aim: Test MNF on patients with chronic steno-occlusive disease, with recurrent transient ischemic attacks (TIA)/minor stroke, presenting regional brain ischemia, at 3 time points (baseline, 8-month and 16-month follow-ups), and comparison with healthy controls.

项目摘要 在这个项目中，我们要开发一种新的非侵入性成像技术，将提供多参数 以前所未有的分辨率绘制了活体大脑的代谢图。这项新技术的关键是新颖性 三个国家的最先进的成像概念的组合：（A）新的硬件，使同时 测量不同频率的多核磁共振（MR）信号;（B）灵活性和鲁棒性 即插即用（PRF）MR指纹识别（MRF）;以及（C）由跨模态模型驱动的数据融合过程 基于统计学习。为了简洁起见，我们将称这种融合的同时多核PnP-MRF技术 多核聚变（Multi-Nuclear Fusion）MNF背后的想法是快速捕获两种不同的定量 （1）来自质子（1H）MRF的结构信息，例如 T1、T2和质子密度（PD）（组织尺度形态）;以及（2）与离子相关的代谢信息 来自钠（23 Na）MRF的稳态，如细胞内钠浓度，以及细胞内、细胞外和细胞外的稳态。 脑脊液（CSF）体积分数（细胞尺度函数）。由于PnP-MRF可以定量多个 组织特性的实验偏差，它使我们能够采用统计学习，以发现一个特定的主题 跨模态模型，其整合了多参数1H PnP-MRF之间的所有体素相互关系 （在高分辨率，0.75-1 mm下获得）和23 Na PnP-MRF（在低分辨率，3-5 mm下获得）图。这些 随后可使用受试者特异性关系来锐化23 Na代谢图以匹配分辨率 1H结构图高分辨率的23 Na地图将使代谢过程的评估， 体内并弥补分辨率上的差距，这些差距阻碍了我们研究活体人脑代谢的能力， 这对我们理解大脑本身以及影响大脑的疾病至关重要。这个概念验证 实施将在7 T下开发，但预期其可适用于临床3 T MR扫描仪。的 具体目标是：（1）数据采集，（1.a）多通道1H/23 Na RF阵列，（1.b）同时多核 3D MRF序列;（2）数据处理，（2.a）1H数据的PnP-MRF重建（指纹匹配 生成结构图）和23 Na数据（组织4室模型和自旋3/2动力学模拟 生成代谢图），（2.b）使用统计学习的跨模态模型，以及数据融合算法， 生成高分辨率代谢图;（3）方法验证，（3.a）准确度和精密度，（3.b）重复性 和再现性。(3)探索性目的：在慢性狭窄闭塞性疾病患者中测试MNF， 在3个时间点（基线， 8-16个月和16个月随访），并与健康对照组进行比较。