Tract-specific statistics based on diffusion-weighted probabilistic tractography

Diffusion-weighted neuroimaging approaches provide rich evidence with which to estimate the structural integrity of white matter in vivo. Tract-based spatial statistics (TBSS), for instance, allows us to assess the spatial distribution of statistical tests performed on diffusion properties such as fractional anisotropy, which are proposed to reflect the integrity of white matter tracts. However, such methods do not provide a direct assessment of white matter integrity for connections between two specific regions of the brain. Here, we present a novel method for deriving tract-specific diffusion statistics, based upon arbitrarily-defined regions of interest. Our approach makes use of an empirically-derived population distribution based on probabilistic tractography, using the Nathan Kline Institute (NKI) Enhanced Rockland sample. We use a heuristic method to determine the most likely geometry of a path between two regions, if one exists, and express this as a spatial distribution. We then estimate the average orientation of streamlines traversing this path, at discrete distances along its trajectory, and compute the fraction of diffusion directed along this orientation for each participant. This allows us to obtain participant-wise metrics along the specific tract (tract-specific anisotropy; TSA), which can then be used to perform statistical analysis on any comparable population. Based on this method, we report both negative and positive associations between age and TSA for two networks derived from published meta-analytic studies (the "default mode" and "what-where" networks), along with more moderate sex differences and age-by-sex interactions in both networks.