Assessment of the performance of SPM analysis in PET neuroactivation studies : A Monte Carlo investigation

Statistical parametric mapping (SPM) is now considered the gold standard both in clinical and research neuroimaging investigations. We investigate the feasibility of Monte Carlo simulations-based assessment of the sensitivity and specificity of SPM analysis in Positron Emission Tomography (PET) neuroactivation studies with respect to design parameters such as study size, activation/lesion localisation, intensity and size. Fifteen data sets were generated for different foci activation levels, and localizations using five spheres located within the digital Hoffman 3D brain phantom with diameters corresponding to 6 and 8 mm, close to the resolution limit of the ECAT ART PET scanner (CTI/Siemens). The five foci were defined in the thalamus (T), putamen (P), cingulate gyrus (CG), right frontal cortex (RFC) and the right supramarginal gyrus (RSG) covering different structures of particular interest in cognitive neuroscience. In addition, we carried out a coordinate search of the resulting maximum cluster, which was found to be close to 10 mm around the true value on the Talairach atlas.