Textural Characterization on Regions of Interest: A Useful Tool for the Study of Small Vessel Disease

We propose a framework for investigating the properties of apparently normal
tissues on brain structural magnetic resonance images of patients with small vessel disease (SVD). It involves the extraction of textural features in regions of interest (ROIs) obtained from an anatomically-relevant template, combined with statistical analysis that considers the relative distribution of SVD markers (e.g. microbleeds, perivascular spaces and white matter hyperintensities) with respect to the ROIs’ textural characteristics in arterial territories derived from another template. We apply this approach to data from 42 patients from a study of mild stroke to investigate whether or not normal tissues in different brain regions are homogeneous regardless of the presence of specific SVD markers and varieties in the manifestations of the pathology (stroke lesion in different arterial territories). Our results suggest that this is not the case: that normal tissues are heterogeneous and that local variations (represented by the entropy) are associated with SVD markers, in agreement with clinical reports.