Normalization in 3D PET: Dependence on the Activity Distribution of the Source

In this communication we show the dependence of the normalization of 3D small animal PET scanners on the calibrating source geometry and non-uniformity of the activity distribution inside it. The potential impact on the reconstructed images of deviations from the ideal activity is discussed. Results from simulated and real data for a pair of block detectors suggest that planar or annular sources can be equally suitable for a normalization procedure if their geometry and their intersections with the system lines-of-response (LORs) are properly taken into account. When comparing an annulus with a centered planar source, both filled with the same total activity, the planar source produces twice the count rate with a 92% of trues versus 66% of trues for the annulus. Scatter and randoms are lower by factors of 3 and 4 respectively in the planar case. We conclude that, although planar and annular sources can be suitable for a normalization procedure if their geometries are properly taken into account, planar source normalization results are better than those of an annulus, particularly if the normalization does not consider the effect of scatter correction, more significant for the annulus source than for the planar one.