Effect of anatomical variability, reconstruction algorithms and scattered photons on the SPM output of brain PET studies

Statistical parametric mapping (SPM) has become the standard technique to statistically evaluate differences between functional images. The aim of this paper was to assess the effect of anatomical variability of skull, the reconstruction algorithm and the scattering of photons in the brain on the output of an SPM analysis of brain PET studies. To this end, Monte Carlo simulation was used to generate suitable PET sinograms and bootstrap techniques were employed to increase the reliability of the conclusions. Activity distribution maps were obtained by segmenting thirty nine T1-weighted magnetic resonance images. Foci were placed on the posterior cingulate cortex (PCC) and the superior temporal cortex (STC) and activation factors ranging between − 25% and + 25% were simulated. Preprocessing of the reconstructed images and statistical analysis were performed using SPM2. Our findings show that intersubject anatomical differences can cause the minimum sample size to increase between 10 and 42% for posterior cingulate Cortex and between 40 and 80% for superior temporal cortex. Ideal scatter correction (ISC) allowed us to diminish the sample size up to 18% and fully 3D reconstruction reduced the minimum sample size between 8 and 33%. Detection sensitivity was higher for hypo-activation than for hyper-activation situations and higher for superior temporal cortex than for posterior cingulate cortex.