Author:

Keywords:

Graph theory, Network analysis, Associative-semantic network

Abstract:

Explicit associative-semantic processing of words and pictures activates a distributed set of brain areas that has been replicated across a wide range of studies. In our lab we have already a long tradition to study these regions in normal subjects as well as in different patient populations. Until recently, these studies were mainly focusing at the regional level, less is known about how information is integrated and transfered between regions. Brain network analysis is a promising tool to investigate cognitive processes. Graph analysis provides a framework to characterize and to quantify networks. Many non-trivial graph characteristics, such as small-worldness, modularity and highly connected hubs, have been observed in human brain networks. The associative-semantic network consists of a large number of nodes and an even exponentially larger number of possible functional connections. Given the extent of the network and the significant lacunae that remain in our knowledge about its internal connectivity structure, we applied graph analysis to characterize and examine the structure of this network. The promises of graph analysis in clinical and basic research to characterize brain connectivity require reproducible and robust results. The reproducibility of graph measures has already been investigated in a number of studies looking at binarized networks derived from structural MRI, diffusion-weighted MRI and resting state fMRI. Only a few studies have looked at the reproducibility of graph measures when using task fMRI. Furthermore, the reproducibility of weighted graph measures has received very little attention: only two studies are available which addressed this problem and both were using graphs derived from diffusion-weighted MRI. Therefore, we investigated the reproducibility and the robustness of graph measures of weighted and binarized networks derived from a task fMRI during explicit associative-semantic processing of words and pictures. Although modeling the same biological system, the human brain, striking differences have been reported. Ways to define node for a connectivity analysis differ. While treating each voxel as a node represents a microscopic view, its high computational burden renders a region-based approach an attractive alternative. Region-based nodes can be task-activated regions or atlas-based regions. An alternative is to apply connectivity-based parcellation to identify functionally homogeneous regions using methods such as Independent component analysis (ICA). However, the difficulty in matching ICA components across subjects and the ignorance of subject specific idiosyncratic effects in group ICA makes this technique difficult to use in network analysis. Recently, a group-wise parcellation schema has been proposed to simultaneously parcellate the group as well as individual network into subunits. The natural correspondence between group and individual subunits makes network analysis among individuals more straightforward. While this relatively new technique for identifying connectivity-based subregions was originally described for resting state functional connectivity, the concept can also be applied to task-based functional connectivity. In comparison of this method to the fully voxel-based approach as well as the previously used task-specific approach, we are able to investigate the granularity effect on the network property. Network analysis has already given insight in many disease studies. In the early phase of a disease, regional differences between healthy subjects and patients may be small while clear changes at the connectivity among regions might already be present. It was shown that dysfunctional effective connectivity, rather than regional changes within brain regions, may contribute to the emergence of language deficits seen in patients with primary progressive aphasia. Comparing associative semantic network properties in patients as well as normal control may give us insight in disease pathology. The aim of this project is to give insight on: (1) Characterizing associative-semantic network in terms of graph metrics; (2) Determining the reproducibility and robustness of the network; (3) Network granularity analysis; and (4) Comparing network properties between elderly normal control and patients.