扩散MRI纤维示踪术的真实精细神经纤维构筑研究

It is well known that the intricate structure of brain is the basis to understand brain functions and neural diseases. Diffusion magnetic resonance imaging (dMRI) tractography provides an in vivo, non-invasive technique for detecting neural fibers in the brain-wide. Based on characterizing the diffusion properties of water molecules in each voxel of several tens of micron, the tracted fibers in dMRI tractography is inferential without knowing its basis of the realistic fiber structure. For a long time, there has been no 3D realistic fine neural fiber strucutre to reconcile this open question. Acquisition of realistic fine neural fiber strucutre is really a great challenge. The invention of micro-optical sectioning tomography (MOST) from the applicants' group has brought unique opportunity to fill this gap. To consolidate the strucutre basis of dMRI tractography, both dMRI and MOST will be applied to acquire the neural fiber structure successively in mice brain ex vivo. Different protocols of dMRI will be used to derive the computational fiber structure in whole brain. MOST will be used to extract the realistic and high spatial resolution structure which is used as standard reference to validate the result of dMRI tractography. An automated and high efficient image processing platform will be developed to deal with the massive data. 3D deformable image registration will be used to align the two kinds of fiber structures for further bidirectional structure-principle validation. We will analyze the correlation and overlapping between the compuational fiber structure and the realistic structure. To mimic dMRI tractography, we will perform mathematical transform of MOST data to do principal component analysis based structure tensor analysis and calculate fiber orientation distribution function.We aim to seek the realistic fiber structure basis of dMRI tractography and optimize the protocol for fiber tracking. By bridging the gap between dMRI tractography and the realistic fiber structure, this proposal will promote the applications of dMRI in basic and clinical studies.

脑的精细解剖结构是理解脑功能和脑疾病的基础。扩散磁共振成像（dMRI）纤维示踪术为检测全脑神经纤维结构提供了活体、非侵入性的手段；但其分辨率低（大于数十微米），以水分子扩散特征间接推算出神经纤维结构。由于缺乏真实神经纤维结构作为参考依据，探究dMRI示踪术与其对应真实纤维结构基础的关系是国际公认难题。申请人团队发展的显微光学切片断层成像（MOST）技术能获取全脑真实精细神经纤维结构，被用来验证dMRI所获取的同源小鼠全脑"计算"纤维结构。本研究拟构建3D海量数据自动处理平台，对真实与"计算"神经纤维结构进行配准融合；首创结构-原理双向推演分析方法：正向分析两类神经纤维的相关性和重合度，反演分析真实纤维结构来模拟dMRI示踪术；发展基于主成分分析的结构张量分析和朝向分布函数分析，优选dMRI示踪术方案。本研究将架起dMRI示踪术与真实神经纤维结构的桥梁，促进dMRI在基础和临床中的应用。

扩散磁共振成像（dMRI）纤维示踪术可为检测全脑神经纤维结构提供活体、非侵入性的手段；但其分辨率低（大于数十微米），以水分子扩散特征间接推算出神经纤维结构。目前，对于dMRI示踪术信号的微观结构基础还不清楚。由于缺乏鼠脑真实神经纤维结构作为参考依据，探究dMRI示踪术与其对应真实纤维结构基础的关系是国际公认难题。本项目中，申请人借助团队前期发展的显微光学切片断层成像（MOST）技术获取小鼠全脑真实精细神经纤维结构，并用于验证dMRI所获取的同品系小鼠全脑“计算”追踪所得纤维结构。本项目主要研究包括：1）构建了一整套dMRI、MOST图像数据获取及3D海量数据处理流程；2）基于先进归一化工具（ANTs）完成dMRI数据和MOST数据的三维非线性配准；3）发展可变尺度主成分结构张量分析并对比分析真实与“计算”神经纤维结构；4）尝试短程轨迹密度成像（stTDI）重建方法并验证其超分辨能力。项目结果研究表明dMRI在解析皮层复杂神经纤维构筑存在固有缺陷，并通过结构张量分析说明了其体素大小依赖性。另一方面，stTDI工作及其验证提示了dMRI重建算法有助于改善dMRI的示踪能力。本项目首创并展示了结构-原理双向推演分析方法：正向分析两类神经纤维在配准之后的相关性和重合度，反演分析真实纤维结构来模拟dMRI示踪术。本项目发展了基于主成分分析的结构张量分析，结合MOST真实神经纤维数据参照，展示了优选dMRI示踪术算法的潜力。本研究可望架起dMRI示踪术与真实神经纤维结构的桥梁，促进dMRI在基础和临床中的应用。