Imaging Water Diffusion in the Brain and in Other Soft Tissues

我们正在继续发明、开发和翻译基于水驱替成像的新型磁共振（MR）方法，从实验台到床边。扩散张量MRI （DT-MRI或DTI）可能是我们发明、开发并成功应用于临床的最著名的成像方法。它测量组织内流动水的扩散张量。它产生的标量参数是组织的固有属性，值得注意的是，我们在不引入造影剂或染料的情况下测量它们。一个dti衍生的量，定向平均扩散系数（或平均ADC），是迄今为止开发的最成功的成像参数，用于可视化进展中的急性中风。随后，我们发现DTI在识别与慢性中风相关的沃勒氏变性方面是有效的。Carlo Pierpaoli及其同事先前对小猫进行的研究表明，DTI在追踪大脑皮层灰质和白质发生的早期发育变化方面很有用，而这些变化是其他成像方法无法检测到的。Sinisa Pajevic和Carlo Pierpaoli开发了一种用颜色编码大脑轴突方向的方法，使我们能够识别和区分结构和组成相似但空间方向不同的白质通路。这些方向编码彩色（DEC）地图清楚地显示了人脑中主要的关联、投射和连接白质通路，是现代神经放射学的支柱。他们甚至可以在诸如《格雷的解剖学》之类的出版物中看到。为了评估大脑不同功能区域之间的解剖连通性，我们还提出并演示了一种使用DTI数据追踪神经纤维束轨迹的方法，为此我们创造了术语DTI“神经束成像”。这是由于Sinisa Pajevic和Akram Aldroubi开发和实现了一个通用数学框架，用于获得测量的离散、噪声、扩散张量场数据的连续、光滑近似，从而使之成为可能。总的来说，这些方法和方法使我们和世界各地的许多其他团队能够对体内的大脑进行详细的解剖和结构分析，这在以前只能使用费力的、侵入性的组织学方法对切除的组织标本进行分析。我们对流线神经束造影的发明和发展的贡献推动了美国国立卫生研究院人类连接组项目（HCP）的创建。

We are continuing to invent, develop, and translate novel Magnetic Resonance (MR) based water displacement imaging methods from the bench to the bedside. Diffusion Tensor MRI (DT-MRI or DTI) is perhaps the best-known imaging method that we invented, developed, and successfully translated clinically. It measures a diffusion tensor of mobile water within tissues. The scalar parameters it produces are intrinsic properties of the tissue and, remarkably, we measure them without introducing contrast agents or dyes. One DTI-derived quantity, the orientationally-averaged diffusion coefficient (or mean ADC), has been the most successful imaging parameter developed to date to visualize an acute stroke in progress. Subsequently, we showed that DTI is effective in identifying Wallerian degeneration often associated with chronic stroke. Previous studies with kittens by Carlo Pierpaoli and colleagues showed DTI to be useful in following early developmental changes occurring in cortical gray and white matter, which are not detectable using other imaging methods. The development of a method to color-encode axon orientation in the brain by Sinisa Pajevic and Carlo Pierpaoli has allowed us to identify and differentiate anatomical white matter pathways that have similar structure and composition, but different spatial orientations. These Direction-Encoded Color (DEC) maps clearly show the main association, projection, and commissural white matter pathways in the human brain, and are a mainstay in modern Neuroradiology. They can even be seen in publications like "Gray's Anatomy". To assess anatomical connectivity between different functional regions in the brain, we also proposed and demonstrated a way to use DTI data to trace out nerve fiber tract trajectories, for which we coined the term DTI "tractography". This was made possible by the development and implementation of a general mathematical framework for obtaining a continuous, smooth approximation to the measured discrete, noisy, diffusion tensor field data by Sinisa Pajevic and Akram Aldroubi. Collectively, these methods and approaches have allowed us and many other groups around the world to perform detailed anatomical and structural analyses of the brain in vivo, which was only possible previously using laborious, invasive histological methods performed on excised tissue specimen. Our contributions to the invention and development of streamline tractography was an impetus for the creation of NIH's Human Connectome Project (HCP). As we began migrating DTI to large, multi-center and multi-patient studies, we began developing a battery of statistical techniques to interpret our imaging data quantitatively, specifically to be able to determine the statistical significance of differences observed in our data. To this end, we developed empirical Monte Carlo and Bootstrap methods for determining features of the statistical distribution of the diffusion tensor from real experimental DTI data. Another innovation was a novel tensor-variate Gaussian distribution that describes the variability of the diffusion tensor in an ideal DTI experiment, and can be used to optimize the design and efficiency of DTI experiments. More recently, we developed approaches to measure uncertainties of many tensor-derived quantities, including the direction of axonal pathways using perturbation and statistical approaches. These developments collectively provide the foundation for applying powerful hypothesis tests to address a wide array of important biological and clinical questions that previously could only be tackled in an ad hoc manner, if at all. More recently, we have been developing sophisticated mathematical/physical models of water diffusion profiles and related these to the MR signals that we measure, with the aim of using our MRI data to "drill down into the voxel" to infer new microstructural and architectural features of tissue (primarilyy white matter in the brain). One example of this is our composite hindered and restricted model of diffusion (CHARMED) MRI framework which provides a mean axon radius for a pack of axons, and an estimate of the intra and extracellular volume fractions. A more recent refinement of CHARMED, AxCaliber MRI, allows us to measure the axon diameter distribution (ADD) within an axon bundle as well from MR displacement imaging data. Sophisticated multiple pulsed field gradient (PFG) NMR and MRI sequences, developed by Michal Komlosh, and translated by Alexandru Avram, help us characterize microscopic anisotropy within tissues like gray matter that are macroscopically isotropic, appearing like a homogeneous gel in DTI. Dr. Komlosh and Ferenc Horkay have developed physical phantoms to test and interrogate mathematical models of water diffusion in complex tissues developed by Evren Ozarslan and Dan Benjamini. Evren also developed novel ways to interpret data obtained from the MR sequences to infer features of size, shape, and distribution of pores in biological tissue and other porous media. He also used the theory of fractals to characterize anomalous diffusion observed in various tissue specimen that are indicative of an underlying hierarchical organizational structure. Collectively, parameters derived from these novel measurements may provide new sources of MR contrast for promising neuroscience applications, such as in vivo (Brodmann or cytoarchitechtonic) parcellation of the cerebral cortex or clinical diagnostic applications, such as improved cancer detection or brain tumor staging. An important development has been a way to characterize non-Gaussian features of the displacement distribution measured using MRI. To this end, our group continues to work on reconstructing the average propagator (displacement distribution) and features derived from it, using a relatively small number of diffusion weighted images (DWI) to enable their clinical migration. The average propagator is the "holy grail" of displacement or diffusion imaging, which can by used to infer geometric features of microscopic restricted compartments as well as glean all of the information provided by DTI as well as other higher-order tensor (HOT) methods. One approach we used previously was an iterative reconstruction scheme along with a priori information and physical constraints to infer the average propagator from DWI data. Another approach was to use CT-like reconstruction method to estimate the displacement profile from DWI data. The most successful method to date, however, developed by Evren Ozarslan, uses special basis functions to represent the average propagator, which compresses the amount of DWI data required while providing a plethora of new imaging parameters or "stains" with which to characterize microstructural features in tissues. A direction our group has pursued more recently is inferring "function" from "structure". This entails using our structural MRI data and inferring function or performance of large-scale brain networks. Alexandru Avram in our group has been developing many of these important applications. Collectively, these novel methods and methodologies represent a pathway to realizing in vivo MRI histology--providing detailed microstructural and microarchitectural information about cells and tissues that otherwise could only be obtained using laborious and invasive histological or pathological techniques applied on biopsied or excised specimens. They also form the core of what is now referred to as "microstructure imaging". We continue to develop new ways to assess tissue structure and architecture in vivo and non-invasively, with the aim of translating these approaches to the clinic, and to the larger biomedical research community.