Intelligent PET/MR Imaging for Mapping Brain Health.

用于绘制大脑健康状况的智能 PET/MR 成像。

• 批准号: RGPIN-2022-05200
• 负责人: Anazodo, Udunna
• 金额: $ 1.82万
• 依托单位: McGill University
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=760705
• 关键词: Intelligent; PET; MR; Imaging; Mapping

PET quantification of brain physiology/neurochemistry are considered gold standards for characterizing brain function and understanding disease pathophysiology in live humans. Integrated positron emission tomography (PET) and magnetic resonance imaging (MRI) scanning, and the increased proliferation of PET probes, have made it more feasible to map fundamental biological mechanisms of the brain and their interactions with the body and the environment. This hybrid technology and the emerging next-generation ultra-high-resolution (UHR) PET imaging systems dedicated to neuroimaging, will permit brain imaging of biological mechanisms at unprecedented spatial precision. However lingering challenges in PET acquisition (high radiation dose), reconstruction (suboptimal attenuation correction), and analysis (invasive kinetic modelling) limit PET's capacity to provide sensitive quantification of molecular changes in the brain, particularly when combined with MRI. Aim- Optimize brain PET/MRI using Intelligent Imaging Solutions: My research program will develop leading-edge intelligent PET/MR imaging (iPET/MRI) tools that leverage advances in machine learning to optimize high-resolution brain PET/MR imaging and address lingering challenges to obtain the best quantitative image quality with the lowest possible radiation dose to then provide new ways of imaging gently (low-dose acquisition, image-derived input functions) and wisely (MRI-guided PET reconstructions), the fundamental mechanisms of complex brain disorders. Approach- Develop iPET/MRI tools using deep learning approaches for Neuroimaging: A prototype iPET/MRI software platform will be developed to amalgamate tools for, 1) optimized ultra-low dose PET acquisition, 2) PET reconstruction for PET/MRI, and 3) PET image analysis to provide a convenient 'one stop shop' tool for neuroscience evaluations. High-fidelity software packages will be first developed first over the next 5 years and validated to provide the foundational components/modules of the iPET/MRI platform. Specifically, deep learning approaches will be used to 1) generate subject-specific high quality PET images from simulated and acquired low-dose data, 2) address gaps in PET/MRI attenuation correction in pediatric populations, 3) provide anatomical-guided (MRI) PET reconstruction (MAP) to improve image contrast and implicitly denoise PET data, and 4) directly generate parametric maps based on Patlak kinetic modelling and image-derive input functions automatically segmented from PET/MRI, eliminating need for serial blood sampling. Feasibility and Impact: Based on my group's recent experience with development of prototype PET analysis software (CALIPER) and rapid application in PET imaging, and access to retrospective PET/MRI data, I am confident that the proposed iPET/MRI will over the next 5 years, produce meaningful use of PET/MRI for neuroimaging, if support from NSERC can be secured to train top talent in computational PET/MRI.

脑生理学/神经化学的PET定量被认为是表征脑功能和了解活体人类疾病病理生理学的金标准。集成的正电子发射断层扫描（PET）和磁共振成像（MRI）扫描，以及PET探针的增加，使得绘制大脑的基本生物学机制及其与身体和环境的相互作用变得更加可行。这种混合技术和新兴的下一代超高分辨率（UHR）PET成像系统专用于神经成像，将允许以前所未有的空间精度对生物机制进行脑成像。然而，PET采集（高辐射剂量）、重建（次优衰减校正）和分析（侵入性动力学建模）中的持续挑战限制了PET提供大脑中分子变化的灵敏定量的能力，特别是当与MRI结合时。 目标-使用智能成像解决方案优化脑部PET/MRI：我的研究计划将开发尖端的智能PET/MR成像（iPET/MRI）工具，利用机器学习的进步来优化高分辨率脑PET/MR成像，并解决挥之不去的挑战，以尽可能低的辐射剂量获得最佳的定量图像质量，然后提供新的成像方式，（低剂量采集，图像衍生的输入功能）和明智的（MRI引导的PET重建），复杂的大脑疾病的基本机制。 方法-使用神经成像的深度学习方法开发iPET/MRI工具：将开发一个原型iPET/MRI软件平台，以合并以下工具：1）优化的超低剂量PET采集，2）PET/MRI的PET重建，以及3）PET图像分析，为神经科学评价提供方便的“一站式”工具。高保真软件包将在未来5年内首先开发，并经过验证，以提供iPET/MRI平台的基础组件/模块。具体而言，深度学习方法将用于：1）根据模拟和采集的低剂量数据生成受试者特定的高质量PET图像，2）解决儿科人群中PET/MRI衰减校正的差距，3）提供解剖引导（MRI）PET重建（MAP）以提高图像对比度并隐式降噪PET数据，以及4）基于Patlak动力学建模和从PET/MRI自动分割的图像导出输入函数直接生成参数图，消除了对连续血液采样的需要。 可行性和影响：基于我的团队最近开发原型PET分析软件（CALIPER）和快速应用于PET成像的经验，以及对回顾性PET/MRI数据的访问，我相信，如果NSERC的支持能够确保培养计算PET/MRI方面的顶尖人才，那么在未来5年内，拟议的iPET/MRI将在神经成像方面产生有意义的PET/MRI应用。