Pseudo-Bootstrap Network Analysis—an Application in Functional Connectivity Fingerprinting

Brain parcellation divides the brain's spatial domain into small regions, which are represented by nodes within the network analysis framework. While template-based parcellations are widely used, the parcels on the template do not necessarily match individual's functional nodes. A new method is developed to overcome the inconsistent network analysis results by by-passing the difficulties of parcellating the brain into functionally meaningful areas. First, roughly equal-sized parcellations are obtained. Second, these random parcellations are applied to individual subjects multiple times and a pseudo-bootstrap of the network is obtained for statistical inferences. It was found that the variation of mean global network metrics from pseudo-bootstrap sampling is smaller compared with inter-subject variation or within-subject variation between two diffusion MRI scans. Using the mean global network metrics from pseudo-bootstrap sampling, the intra-class correlation is always higher than the average obtained from using a single random parcellation. As one application, the pseudo-bootstrap method was tested on the Human Connectome Project resting state dataset to identify individuals across scan sessions based on the mean functional connectivity – a trivial network property that has little information about the connectivity between nodes. An accuracy rate of ~90% was achieved by simply finding the maximum correlation of mean functional connectivity of pseudo-bootstrap samples between two scan sessions.