Comparison of energy window choice and parameter implementation in dual energy window scatter correction performance in 3D PET

The authors have investigated the relevance of the energy window choice in the dual energy window (DEW) scatter correction method (Grootoenk et al., 1991). This method depends strongly on the difference between 2 parameters that need to be estimated from phantom studies, the ratio (R/sub unsc/) between the true events in the low energy window (LEW) and high energy window (HEW), and the equivalent ratio for the scattered events (R/sub sc/). The transaxial distribution of the R/sub unsc/ is relatively flat regardless of the energy window selection. However, the energy window choice considerably influences the transaxial distribution of R/sub sc/. Since the DEW scatter correction method does not make use of convolution techniques, spatial variations of parameters are expected to lead to nonuniformities in scatter corrected images. The authors have investigated the DEW scatter correction accuracy and uniformity as a function of energy window choice and determination of R/sub sc/. Based on stability considerations 2 sets of energy windows have been selected for study, 100-350-850 keV and 250-380-850 keV. Three methods to implement the R/sub sc/ values in the scatter correction estimate have been used: an array containing the transaxial distribution of R/sub sc/; a constant value obtained by averaging R/sub sc/, over the central 20 cm diameter region; and a constant value corresponding to the central R/sub sc/ value. The use of constant values for R/sub sc/ improved image uniformity for both sets of energy windows from within 30% obtained when the array for R/sub sc/ was used, to about 15%. The uniformity was estimated over a 20 cm diameter region. Scatter correction accuracy as determined from uniform phantom and cold spot studies was comparable in all cases. No significant difference in transaxial or axial image uniformity was observed between data obtained with the 2 sets of energy windows and normalized with the present normalization algorithm. Likewise the accuracy obtained in both cases was comparable, indicating that the implementation method of the R/sub sc/ is a more important factor in the DEW scatter correction method performance than the selection of energy windows.