Connectomics can be used to investigate functional brain networks in neurodegenerative diseases including Huntington's disease (HD). In this developing field, different connectome construction strategies have emerged in parallel. However, there is a need to understand the influences of different strategies on subsequent analyses when constructing a connectome. This study systematically compares connectome construction strategies based on their biological relevance to functional networks in neurodegeneration. We asked which functional connectome construction strategy was best able to discriminate HD gene carriers from healthy controls, and how such a strategy affected modular organization of the network. The major factors compared were principal component-based correction versus wavelet decomposition for physiological noise correction, the type of parcellation atlas (functional, structural and multi-modal), weighted versus binarized networks, and unthresholded versus proportionally thresholded networks. We found that principal component-based correction generated the most discriminatory connectomes, while binarization and proportional thresholding did not increase discrimination between HD gene carriers and healthy controls. When a functional parcellation atlas was used, the highest discrimination rates were obtained. We observed that the group differences in modular organization of the functional connectome were greatly affected by binarization and thresholding, showing no consistent pattern of modularity. This study suggests that functional connectome construction strategies using principal component-based correction and weighted unthresholded connectivity matrices may outperform other strategies.

Assessment of Functional Connectome Construction Strategies in Neurodegeneration