The Effect of Parcellation, Tract Number and Network Characteristic for Brain Network Analysis

• Main Authors: Yun Ting Ciou, 邱筠婷
• Other Authors: Y. P. Chao
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/8ypg4w

碩士 === 長庚大學 === 資訊工程學系 === 105 === The brain works achieving complex cognitive functions have to depend on anatomical and functional connectivity which are formed into brain network. Therefore, many research combine diffusion MRI with tractography using graph theory to describe graph theoretical modeling of brain connectivity. This model comprises the cortical/sub-cortical regions represented as nodes and the connection between pairs of regions represented as edge. It finds out the “small-world” properties of brain networks and non-random composite of router. Besides, the parameter which is used to describe network properties in graph theory such as degree and local efficiency is associated with external behavior performances, cognitive functions, diseases, development of nerve and brain plasticity. However, the network structure and the network property will change when using distinct nodes and edges. This study used before and after diffusion MRI images from two groups’ (Adaptive group and non-adaptive group) training through different level and the performance in working memory tasks. Under the premise that we knew learning will affect working memory abilities, this paper assumed the brain structure will change by neuroplasticity arose from learning and used two kinds of diffusion parameters-fractional anisotropy and restricted volume fraction, ten numbers of cortical/sub-cortical regions, four numbers of fiber tracts, eleven numbers of binary threshold, and seven kinds of parameter which is used to describe network properties so as to investigate what combination could observe the biggest difference between two groups in network properties and understand what parameter could reflect the cognitive performance related to working memory. The results show that fractional anisotropy could be more easier to distinguish between two groups. When the fiber tract number was one-eighth of total neural fibers tracts (about 125,000) and the cortical parcellation number was 180, 540, 630, 810, 900, 990, the result was significant (p<0.05) by using normalized characteristic path length, characteristic path length and small-worldness of network properties. In addition, it was more significant if the binary threshold of fiber tracts higher than 10. When using restricted volume fraction, it only could distinguish between two groups in 450 or 630 cortical parcellations with small-worldness of network properties. In the other hand, the correlation between working memory and network properties which was disparity in two groups was significant. When using fractional anisotropy, the results show that the correlation between forward digit span/backward digit span and normalized characteristic path length/ characteristic path length was significant (-0.44<r<-0.236, p<0.05). When using restricted volume fraction, the results show that the correlation between forward digit span and small-worldness was significant (-0.516<r<-0.488, p<0.05). Besides, these significant correlations were only found in non-adaptive group, but not found in adaptive group. In this paper, different process of learning will vary the brain structure changing. It can be observed these change by MRI and the network property of graph theory, but nonetheless not all of defined nodes and edges can reflect the difference. These findings indicate that three kinds of network properties, normalized characteristic path length, characteristic path length and small-worldness, are more sensitive to detect the brain structure changing caused by learning. The change between two groups in 630 cortical parcellations by means of both diffusion parameters is also can observe. It is suggested that future studies could be applied to other types of studies, and use further more quantity of cortical regions or different kinds of diffusion parameters and tractography to analysis.