Selection of Parcellation Atlas in DTI Brain Connectome Analysis

Analysis of the structural connectomes based on diffusion magnetic resonance imaging (MRI) can show network attributes for segregation and integration in the human brain. However, it is reported that there are large discrepancies from those network measures, which may result from the network construction methodology (such as brain parcellations, DWI acquisition model, tractography algorithms and network weighting schemes). The influence of parcellation atlas on the topological network measures of diffusion tensor imaging (DTI) network with the typical setting used in many clinical applications were evaluated in this paper. The data of research were acquired in 75 healthy older adults who underwent several neuropsychological tests and structural MRI. It shows that those network measures are very sensitive to the spatial scales of brain parcellation. When the scale is close, the principle of parcellation doesn't change those measures that much. The modularity of brain network is relatively stable on different kinds of brain parcellation. The mesoscale atlas, which shows higher sensitivity in ageing study, may be an optimal choice for typical DTI network analysis.