磁共振扩散张量成像数据的统一分析框架及应用研究

Diffusion Tensor Imaging (DTI) is a rapidly developing new MRI modality for brain white matter imaging in vivo, being able to reveal the structure and integrity of the underlying white matter fiber tracts. Thus, it has a wide range of applications in normal brain development, aging, neuropsychiatric disorders and cognitive research areas. Currently, there are no widely recognized standard DTI analysis methods and software tools for large scale clinical DTI researches.Different analysis methods can give nonconherent, even conflict results. Due to the intrinsic low SNR (signal to noise ratio) and long scanning time, DTI data collected from clinical trials inevitably suffers from several kinds of artifacts and defects. Group wise contrast studies always prefer an unbiased DTI atlas with high accuracy. This means all the subjects should be accurately registered into a common coordinates bearing accurate spatial correspondence. However, currently wide used DTI group wise analysis tool, such as SPM and TBSS, are far from perfect for clinical studies with hundreds of DTI data sets in terms of quality control, unbiased atlas computing and robust statistical analysis.This greatly weaken the power of DTI to draw solid conclusions in neuropsychiatric studies. This project aims to set up a unified framework for DTI data analyses and use it to analyze first-onset schizophrenia.Based on our previous work on DTI, we plan to conduct detailed research on DWI quality control based on the evenness measurement of directional distribution of diffusion measuring gradients in quality controlled DWI data sets, DTI quality control based on regional DTI principal direction distribution entropy, a two step DTI registration strategy combining the scalar based and whole tensor based DTI registration methods, principal direction distribution entropy along fiber tracts for group wise fiber tract based analysis method. We expect to improve both the quality and computational cost of group wise DTI study of clinical experiments. The proposed methods will be used to evaluate the initial-onset schizophrenia group against the normal controls and other clinical neuropsychiatric disorders. This research will definitely facilitate large scale clinical DTI research with more robust pipeline and tools.

磁共振扩散张量成像（DTI）是目前唯一的活体脑白质纤维束成像方法。它能够测量白质纤维束的走行方向和细微结构完整性，在脑发育、神经精神疾病和认知等研究领域有广范的应用。现有的几种DTI分析方法在DWI和DTI 的自动质量控制、DTI无偏图谱的计算和DTI 的统计对照分析等方面均不够完善，造成了一些临床研究的报道出现不一致、甚至互相矛盾的结果。因此，迫切需要把各种分析方法纳入统一的分析框架，使得各种分析的结果可以互相比较和印证。基于我们已有的DTI 研究基础，本项目以首发精神分裂症为研究背景，研究并构建统一的DTI数据分析框架。重点研究剔除部分劣质DWI梯度数据后DWI数据集的可用性、基于区域内张量主轴方向分布熵测度的DTI自动质量控制、结合标量和张量图像的多通道非线性DTI图像配准方法、成组DTI图像样本无偏图谱的快速、精确构建和沿纤维束的张量主方向分布熵测度的纤维束对照分析等新的分析方法。

磁共振扩散张量成像（DTI）是目前唯一的活体脑白质纤维束成像方法。它能够测量白质纤维束的走行方向和细微结构完整性，在脑发育、神经精神疾病和认知等研究领域有广范的应用。本项目以视神经脊髓炎（NMO）为研究背景，主要研究内容包括DTI/DWI数据的自动化质量控制、基于张量信息的DTI图谱计算和基于图谱的DTI数据自动分析等。DWI/DTI质量控制方法研究方面，继续开发和改进DTIPrep (http://www.nitrc. org/projects/dtiprep)工具软件，优化了DTIPrep软件的自动运行控制参数，参数使得DTIPrep成为DTI研究领域里一个有力的工具软件。DTI无偏图谱计算方面，使用基于张量信息的算法代替FSL等主流DTI分析工具包中基于标量信息驱动的图像配准方法。在FSL的TBSS流程中成功替换使用DTITK配准步骤，改进了图谱计算的质量，既优化改进了TBSS算法的结果，又可以提供更好地基于体素的分析结果。为在图谱空间的基于张量的结构化脑网络构建与分析提供了更好的方法和更加可信的结果。