An examination of the statistical reliability and prognostic potential of a measure of 3D spatial heterogeneity developed for volumetric PET images of human sarcomas

Local heterogeneity is an important descriptor of the biological behavior of normal and pathological tissue. The most widely used measure of heterogeneity is relative dispersion(RD)(Kendal, PNAS, 2001). Because relative dispersion is spatially invariant it cannot distinguish alternative spatial patterns in the structure of heterogeneity. In previous work with FDG PET studies of human sarcomas a measure of spatial heterogeneity was developed that is based on the deviation of the distribution of the standardized FDG uptake (SUV image) from a unimodal elliptically contoured spatial pattern(O´Sullivan et al, SNM, 2001). This paper reports studies carried out to evaluate the statistical reliability and performance characteristics of this heterogeneity measure.