Using Classical Geostatistics to Quantify the Spatiotemporal Dynamics of a Neurodegenerative Disease from Brain MRI

Abstract

We present a novel approach to characterize multiple sclerosis (MS) from brain magnetic resonance imaging (MRI) with geostatistics. Brain MRI provides excellent, exhaustive input data to geostatistical analysis, typically several million voxels per MRI scan. A dataset of 259 spatially normalized binary MS white matter lesion (WML) patterns covering very mild to extremely severe MS cases was subject to directional variography. Using an exponential variogram model function, the observed spatial variability in x, y, z directions can be expressed by geostatistical parameters range and sill which perfectly correlate with WML pattern surface complexity and lesion volume. A scatterplot of ln(range) vs. ln(sill), classified by pattern anisotropy, enables a consistent and clearly arranged presentation of MS-lesion patterns based on their geometry. The geostatistical approach and the graphical representation of results are considered efficient exploratory data analysis tools for longitudinal, cross-sectional, and medication impact studies.