Zhu Dajiang, Zhang Degang, Faraco Carlos, Li Kaiming, Deng Fan, Chen Hanbo, Jiang Xi, Guo Lei, Miller L Stephen, Liu Tianming
University of Georgia, Dept. of Computer Science, 415 Boyd GSRC Building, Athens, GA, USA.
Inf Process Med Imaging. 2011;22:97-110. doi: 10.1007/978-3-642-22092-0_9.
The lack of consistent and reliable functionally meaningful landmarks in the brain has significantly hampered the advancement of brain imaging studies. In this paper, we use white matter fiber connectivity patterns, obtained from diffusion tensor imaging (DTI) data, as predictors of brain function, and to discover a dense, reliable and consistent map of brain landmarks within and across individuals. The general principles and our strategies are as follows. 1) Each brain landmark should have consistent structural fiber connectivity pattern across a group of subjects. We will quantitatively measure the similarity of the fiber bundles emanating from the corresponding landmarks via a novel trace-map approach, and then optimize the locations of these landmarks by maximizing the group-wise consistency of the shape patterns of emanating fiber bundles. 2) The landmark map should be dense and distributed all over major functional brain regions. We will initialize a dense and regular grid map of approximately 2000 landmarks that cover the whole brains in different subjects via linear brain image registration. 3) The dense map of brain landmarks should be reproducible and predictable in different datasets of various subject populations. The approaches and results in the above two steps are evaluated and validated via reproducibility studies. The dense map of brain landmarks can be reliably and accurately replicated in a new DTI dataset such that the landmark map can be used as a predictive model. Our experiments show promising results, and a subset of the discovered landmarks are validated via task-based fMRI.
大脑中缺乏一致且可靠的具有功能意义的标志物，这严重阻碍了脑成像研究的进展。在本文中，我们使用从扩散张量成像（DTI）数据中获得的白质纤维连接模式作为脑功能的预测指标，并在个体内部和个体之间发现一个密集、可靠且一致的脑标志物图谱。一般原则和我们的策略如下。1）每个脑标志物在一组受试者中应具有一致的结构纤维连接模式。我们将通过一种新颖的迹线映射方法定量测量从相应标志物发出的纤维束的相似性，然后通过最大化发出纤维束形状模式的组内一致性来优化这些标志物的位置。2）标志物图谱应密集且分布在大脑的所有主要功能区域。我们将通过线性脑图像配准初始化一个包含约2000个标志物的密集且规则的网格图谱，这些标志物覆盖不同受试者的整个大脑。3）脑标志物的密集图谱在不同受试者群体的不同数据集中应具有可重复性和可预测性。通过可重复性研究对上述两个步骤中的方法和结果进行评估和验证。脑标志物的密集图谱可以在新的DTI数据集中可靠且准确地复制，从而使标志物图谱可以用作预测模型。我们的实验显示出了有前景的结果，并且通过基于任务的功能磁共振成像对发现的一部分标志物进行了验证。
