Greg D Parker1, Mark Drakesmith1,2, and Derek K Jones1,2
1CUBRIC, School of Psychology, Cardiff University, Cardiff, United Kingdom, 2Neuroscience and Mental Health Research Institute (NMHRI), School of Medicine, Cardiff University, Cardiff, United Kingdom
Graph
theoretical connectome analysis1
is an increasingly important
research area. There is, however, high
computational overhead required to: (a) produce whole or partial brain
tractographies; (b) convert tractographies into binary or weighted
graphs; and (c) analyse those graphs according to multiple, often
complex graph metrics. We have developed GP-GPU
accelerated implementations of each step. Exploiting the resultant
increase in computational power, we examined the effects of increasing
streamline sampling
densities and number of cortical parcellations
on separability of connectomes between first episode psychosis patients and controls. We show finer cortical parcellation increases separability (while increasing streamline density reduces it).
