Comparison analysis of diffusion-tensor-imaging tractography in native space and in normalized space: applied for brain network analysis based on human brain atlas

Abstract:
Brain network analysis has been widely used to study the development of the brain and various neural diseases. Brain networks are mainly constructed based on functional Magnetic Resonance Imaging (fMRI): resting state Blood-Oxygen-Level-Dependent (rs-BOLD) and Diffusion-Weighted-Imaging (DWI) or Diffusion-Tensor-Imaging (DTI) signals. The nodes of the network are normally defined by ROIs, typically brain atlases labelled in normalized space, whereas the edges are defined as the connectivity between each two nodes. If the connectivity is defined as the time-domain correlation between rs-BOLD data of two nodes, it is called functional connection network (FCN). If the connectivity is defined as the number of fibers which represent the spatial connections between two nodes, it is called diffusion tractography network or structural connection network (SCN). The SCN may vary depending on tractography space selected. We call each person’s space, such as the original DTI volume space, the native space. Brain template space, such as the template ICBM152 in which the brain atlas is labelled, is called the normalized space. This study aims to evaluate which space is more suitable for constructing brain SCN, especially for clinic application. A comparison analysis experiment of tractography space is presented in this paper. The correlation between SCN and FCN is used as the comparison criteria. We use two brain atlases (Brodmann and AAL116), three connectivity networks (one FCN in normalized space, two SCNs with one in native space and the other in normalized space), and three datasets (healthy subjects, spinal cord injury patients and stroke patients) in this study. Both the correlation between FCN and SCN, and the t test between healthy subjects and patients’ datasets are used for evaluating tractography space. The results show a positive correlation between FCN and SCN. The correlation coefficients between FCN and SCN in native space are significantly greater than that in normalized space. This inspires us that tractography in native space would result in higher correspondence with the functional organization. The independent t test of the connection strength between “Healthy” and “SCI patients”, “Healthy” and “Stroke” shows that more significant difference of connection strength will be explored by the SCN in native than in normalized space. It is verified that the tractography in native space should be a better choice when studying brain connectivity in clinic application.