Markov Chain Monte Carlo Random Effects Modelling in Diffusion MRI: a New Window on Microstructure and White Matter Architecture

磁共振成像(MRI)彻底改变了我们制作大脑图片的方式。核磁共振成像可以通过测量水在脑组织微观结构中的移动方式来产生大脑结构的图像。大脑的白质由连接不同大脑区域的长细胞结构组成，这可以被认为是大脑的线路。因为水可以更容易地沿着这些结构移动，所以有可能用一种名为气管造影术的技术来产生这些连接的图像。轨迹造影术在许多临床研究中很有用，尤其是在神经外科计划中。这位神经外科医生的目标是切除大脑中不正常的部分，同时保持这些重要的连接完好无损。因此，纤维束造影术通过显示与要切除的异常相关的重要路径的位置，帮助神经外科医生导航。因此，这些技术可以帮助改善患者的预后，减少手术后的残疾。然而，应用这些技术的挑战是，它们必须在有限的MRI信息下使用，因为病人无法忍受长时间的扫描。核磁共振扫描是数字化的，就像数码相机的图像一样，由图片元素或像素组成。我们将这些称为体素，因为MRI扫描具有相关的切片厚度。MRI扫描的分析通常是在逐个体素的基础上完成的。因此，该常用方法将每个体素中的信号视为独立的。在这个项目中，我们将使用一种方法，这种方法不做这种假设，而是使用来自相邻或类似响应的体素的信息。我们将这种方法称为信息借用，并使用贝叶斯随机效应模型来实现这一点。我们的目标是应用这种方法来提高纤维束造影术的准确性和稳健性，从而为临床医生提供更好的工具来进行神经外科计划和其他临床研究。我们还将把这些方法应用于测量大脑微结构的新技术中。后一个目标虽然雄心勃勃，但将提供新的组织结构标记，更直接地代表负责大脑功能的结构。

Magnetic resonance imaging (MRI) has revolutionised the way in which we can produce pictures of the brain. MRI can be used to produce pictures of the brain structure by measuring the way in which water can move around in the microscopic structure of the brain tissue.The white matter of the brain consists of long cellular structures that connect different brain regions togther, this can be thought of as the wiring of the brain. Because the water can move more easily along these structures, it is possible to produce images of these connections with a technique called tractography.Tractography is useful in many clinical investigations, most notably in neurosurgical planning. The neurosurgeon aims to remove an abnormal part of the brain whilst leaving these important connections intact. Tractography therefore helps to navigate the neurosurgeon by revealing the location of important pathways with respect to the abnormality to be removed. These techniques can therefore help to improve patient outcomes and reduce post-surgical disability.The challenge however, in applying these techniques is that they have to be used with limited MRI information, because sick patients are unable to tolerate long scans. An MRI scan is digitized and just like an image from a digital camera consists of picture elements or pixels. We refer to these as voxels because the MRI scan has an associated slice thickness.Analysis of MRI scans is usually done on a voxel by voxel basis. This commonly used method therefore treats the signal in each voxel as independant. In this project we will use a method that does not make this assumption but uses information from adjacent or similarly responding voxels. We call this approach information borrowing and achieve this using Bayesian random effects modelling.We aim to apply this approach to improve the accuracy and robustness of tractography thereby providing clinicians with better tools for neurosurgical planning and other clinical investigations. We will also apply these methods in new techniques for measuring the brain microstructure. This latter aim, although ambitious, will provide new markers of tissue structure that are more directly representative of the structure responsible for the functioning of the brain.