Investigation of the spatial correlation in human white matter and the influence of age using 3-dimensional variography applied to MP-RAGE data.

ABSTRACT

A novel method for the quantification of heterogeneity and spatial correlation in 3D MP-RAGE images of white matter is presented. The technique is based on the variogram, a tool commonly used in geosciences for the analysis of spatial data, and was tailored to the special requirements of MR image analysis. Influences from intensity non-uniformities, noise and arbitrary greyscale were quantified and considered in the calculations. The obtained variograms were fitted with spherical model functions to infer parameters that quantify heterogeneity and size of the correlation structures of the tissue. Numerically generated samples with well-defined correlation properties were employed to validate the estimation process and to provide an interpretation of the parameters obtained. It is shown that the method gives reliable results in an interval of correlation structures sized between 2mm and 20mm. The method was applied to 24 MP-RAGE datasets of healthy female volunteers ranging in age from 19 to 73 years. White matter was found to have two prominent correlation structures with sizes of approximately 3mm and 23 mm. The heterogeneity of the smaller structure increases significantly with age (r=0.83, p<10(-6)).