Exploratory Analysis Tools for Developmental Studies of Brain Microstructure with Diffusion MRI

项目摘要 弥散MRI分析是研究脑白色物质显微结构的主要工具。青少年大脑 由于大脑重组和神经修剪，发育涉及显著的微观结构变化。的 这些变化的轨迹可以通过精神障碍的发作而改变，并且纵向分析 需要扩散成像来捕获这些改变。我们的建议旨在提供一个开放源代码 纵向弥散MRI数据集探索性分析的软件平台，并将其应用于研究 青少年脑认知发育（ABCD）数据中双相情感障碍的早期发作。 扩散张量成像不足以区分不同类型的变化，发生在 青少年大脑发育，所以我们的软件工具将支持更先进的扩散模型， 神经突取向分散和密度成像。这利用了 ABCD扩散图像。我们开发的探索工具将采用广泛使用的基于区域的空间 统计（TBSS）技术，这是一种非常适合探索性分析的降维方法。 超越标准TBSS，我们将调整该技术，以与改进的配准技术配合使用， 使用更先进的扩散模型，并提供基于纵向 成像数据。 TBSS方法需要高质量的图像对准。为了实现这一目标，我们将制定一项 深度学习模型用于扩散数据集的成对配准，加速扩散的重要一步 MRI分析。通过从ABCD扩散成像生成纤维取向分布，我们的配准 除了解剖结构外，模型还将对齐纤维方向。 这些工具将在一个开放源码软件平台上提供。虽然有许多工具， 探索扩散MRI数据，这一个将是独特的，在其支持纵向数据集与高角度 分辨率成像

Project Summary Diffusion MRI analysis is the primary tool for studying brain white matter microstructure. Adolescent brain development involves significant microstructural changes due to brain reorganization and neural pruning. The trajectory of these changes can be altered by the onset of mental disorders, and longitudinal analysis of diffusion imaging is needed to capture these alterations. Our proposal aims to make available an open source software platform for exploratory analysis of longitudinal diffusion MRI datasets, and to apply it to study the early onset of bipolar disorder in the Adolescent Brain Cognitive Development (ABCD) data. Diffusion tensor imaging is not sufficient to distinguish the different types of changes that occur during adolescent brain development, so our software tool will support more advanced diffusion models such as Neurite Orientation Dispersion and Density Imaging. This takes advantage of the high angular resolution of ABCD diffusion images. The exploratory tools we develop will employ the widely used Tract-Based Spatial Statistics (TBSS) technique, which is a dimensionality reduction that is well suited for exploratory analysis. Going beyond standard TBSS, we will adapt the technique to work with improved registration techniques, to work with more advanced diffusion models, and to provide developmental trajectories based on longitudinal imaging data. The TBSS methodology requires high quality image alignment. In order to achieve this, we will develop a deep learning model for pairwise registration of diffusion datasets, speeding up an important step in diffusion MRI analysis. By generating fiber orientation distributions from the ABCD diffusion imaging, our registration model will align fiber orientations in addition to anatomical structures. These tools will be made available in an open source software platform. While there are many tools for exploring diffusion MRI data, this one will be unique in its support for longitudinal datasets with high angular resolution imaging.