Background-adaptive dual-energy-window correction for Compton scattering in SPECT

Detection of gamma rays which Compton scatter within a patient but are still within the photopeak window of the Anger camera leads to inaccuracies in quantification of radioactivity from nuclear-medicine images. With the dual-energy-window correction method and a single (universal) scatter multiplier, the activity error is relatively large when there is tissue background with a large range of values including zero . We examine here a procedure that adapts the scatter multiplier to the level of background . In a Monte Carlo investigation, we introduce the iterative technique, examine when it converges, and look at the resultant improvement in quantification . Three geometries are checked: a large-sphere 99'Tc target within a cylinder containing 1) a uniform or 2) non-uniform background and 3) a 1231 brain phantom. Reconstruction of the data is carried out with the iterative maximum-likelihood, expectation-maximization algorithm with attenuation correction . Results show that the multiplier converges to a stable value after only a few iterations for all cases. Typical errors in target activity are : for the off axis sphere in non-uniform background 23.3% (no correction), -13 .9% (universal-multiplier correction), and -0.7% (converged-multiplier correction); for the putamen in uniform white-matter background 20.4% (no correction), 10.6% (universal-multiplier correction), and 3.0% (converged-multiplier correction) . The iterative background-adaptive method leads to considerable improvement in all cases tested . In general, detection of gamma rays which Compton scatter within a patient but are still within the photopeak window of the Anger camera leads to inaccuracies in quantification of radioactivity from nuclear-medicine single-photon emission computed tomography (SPECT). The dual-energy-window correction method [1,2], in which counts in a lower-energy window are multiplied by a "scatter multiplier" and subtracted from those in the photopeak window, is simple, and Monte Carlo investigations indicate that in 99mTc SPECT of focal, radioactive regions, correct activities can be obtained . A single (universal) scatter multiplier is fairly accurate in the face of variations in patient size and in location of the focal activity . However, the activity error is much larger for the cases where variation in tissue background has a large range of values including zero [3,4] . We examine here a procedure that adapts the scatter multiplier to the level of background. In a Monte Carlo 0168-9002/94/$07.00 C 1994 Elsevier Science B.V . All rights reserved SSD10168-9002(94)00769-1 NUCLEAR INSTRUMENTS & METHODS IN PHYSICS RESEARCH SectionA investigation, we 1) introduce the iterative technique, 2) examine when it converges, and 3) look at the resultant improvement in quantification . Complications from high levels of statistical noise are expected to be small but are not considered .