Statistics and Its Interface
Volume 3 (2010)
Nonparametric evaluation of heterogeneity of brain regions in neuroreceptor mapping applications
Pages: 59 – 67
In many applications of modeling positron emission tomography (PET) data for neuroreceptor mapping studies, it is necessary to define one or more regions of interest (ROI) and analyze aggregate data from each region, often assumed to be homogenous. We propose a simple method for assessing the level of heterogeneity within any given ROI along with a procedure for testing a null hypothesis of regional homogeneity that uses a wild bootstrap algorithm. Estimation of outcome measures is accomplished using a mixture modeling approach. We provide results of a simulation study along with analysis of an imaging dataset, which indicates that most of the ROIs considered are quite heterogeneous.
wild bootstrap, mixture modeling, positron emission tomography, compartment modeling