An investigation of practical scatter correction techniques for 3D PET

Abstract

The operation of septa-retractable multi-ring PET scanners in 3D mode results not only in a six-fold increase in sensitivity, but also in a three-fold increase in the background due to scattered photons. In PET applications where accurate quantitation is required, an effective scatter correction technique is necessary. The goals is to correct scatter to a level at least comparable with standard 2D PET, where many of the scattered photons are eliminated by lead septa. Recently, a number of scatter correction techniques have been proposed, including a dual-energy window subtraction, a convolution-subtraction method using both a constant and a spatially-dependent scatter fraction, a model-based approach where the mean number of scatters is directly calculated, and others. In order to investigate the efficacy of these different techniques, a new phantom has been designed and built with a high degree of inhomogeneity both transaxially and axially. The phantom consists of a cylinder 20 cm in diameter and 15 cm in length, containing four compartments: (1) an annulus 15 cm long, 2 cm thick and with an outside diameter of 20 cm, (2) an inner cylinder 15 cm long and 18 cm in diameter, (3) a small cylinder 10.5 cm long and 4.5 cmmore » diameter and (4) a shorter cylinder 5.5 cm long and 4.5 cm diameter. A fifth cylindrical compartment, 10 cm long and 20 cm in diameter is attached to one end of the phantom to provide out-of-field activity when required. A series of studies has been performed with different activity concentrations in the five compartments, and data acquired in both 2D and 3D with an ECAT 951R septa-retractable scanner. All 3D data sets have been collected in dual energy window mode, and scatter-corrected using either the DEW or CON technique prior to reconstruction.« less