Does resolution modeling during reconstruction improve the DAT measurement on a high resolution PET?

Objectives Resolution modeling is known to improve quantification in small structures by reducing the partial-volume effects. The aim of this work is to assess the impact of a fast reconstruction including resolution modeling on the BP_ND of ¹¹C-PE2I to DAT using the high resolution HRRT PET scanner (Siemens) in Human.

Methods Fourteen healthy subjects were examined during one hour after injection of ¹¹C-PE2I. ROI were delineated in striatum and cerebellum using T1 MRI images. Each PET acquisition was reconstructed twice using the ordinary Poisson OSEM algorithm with and without implementation of the resolution modeling during reconstruction. The number of OSEM iterations was determined such as to ensure convergence of the BP_ND in the striatum. The BP_ND was calculated in striatum using ROI and simplified method. Voxel-based BP_ND parametric maps were consecutively generated and normalized using SPM5 and 11C-PE2I templates. Paired t-tests between reconstructions were performed on ROI (p<0.05) and voxel-based (SPM5; p<0.05, FDR) approaches.

Results 12 and 20 OSEM iterations (16 subsets) were chosen respectively without and with resolution modeling. Resolution modeling results in a 20 +/- 6 % BP_ND increase in the dorsal striatum and a 14 +/- 3.5 % in the ventral striatum. The SPM analysis shows significant increase of BP_ND in the dorsal and ventral striatum, the midbrain and the parahippocampal area. No difference in DAT-free regions was observed.

Conclusions Resolution modeling during reconstruction increases consistently the BP_ND values in DAT-rich regions. This increase should allow the detection of small changes of biological parameters for clinical purpose.