The aim of this study was to demonstrate the necessity of an off-centre uniformity measurement during performance evaluation and acceptance testing of a positron emission tomography (PET) system. To this end, the effect of different methods of geometric arc correction on image uniformity was considered. The arc correction routine of the system software of a particular PET scanner family was tested in computer simulations, phantom measurements and a patient study. Various methods of geometric arc correction--nearest neighbour interpolation, linear interpolation and cubic smoothing spline interpolation--were applied to the same data. Uniformity was evaluated both visually and quantitatively using intensity profiles and regions of interest. The arc correction routine of the PET scanner family produced significant ring artefacts and led to overestimation of tracer uptake by up to 15%. Since uniformity measurements are usually performed using a cylindrical phantom at the centre of the transverse field of view, these artefacts are not detected. In conclusion, the standards for performance evaluation of a PET scanner should be extended by inclusion of an off-centre uniformity measurement at the edge of the transverse field of view. On the basis of our comparison of different methods for geometric arc correction, we suppose that cubic smoothing spline interpolation might improve the relation between statistical noise reduction and spatial resolution as compared with conventional linear interpolation.
