Stability and sensitivity of structural connectomes: effect of thresholding and filtering and demonstration in neurodegeneration

Abstract

Structural connectomes derived using diffusion tractography are increasingly used to investigate white matter connectivity in neurological diseases. However inherent biases in diffusion tractography algorithms may lead to both false negatives and false positives in connectome construction. A range of graph thresholding approaches and more recently several streamline filtering algorithms have been developed to address these issues. However there is no consensus in the literature regarding the best available approach. Using a cohort of Huntingtons disease patients and healthy controls we compared the effect of several graph thresholding strategies: proportional, absolute, consensus and consistency thresholding, with and without streamline filtering, using Spherical Deconvolution Informed Filtering of tractograms (SIFT2) algorithm. We examined the effect of thresholding strategies on the stability of graph theory metrics and the sensitivity of these measures in neurodegeneration. We show that while a number of graph thresholding procedures result in stable metrics across thresholds, the detection of group differences is highly variable. We also showed that the application of streamline filtering using SIFT2 resultes in better detection of group differences and stronger clinical correlations. We therefore conclude that the application of SIFT2 streamline filtering without graph thresholding may be sufficient for structural connectome construction.