Evaluation of the effect of scatter correction on lesion detection in hepatic SPECT imaging

Various methods have been proposed to correct for the problems presented by scatter in images from single-photon computed tomography (SPECT). While improvement in quantitative accuracy has been reported with a variety of measures, the effect on the accuracy of lesion detection requires analysis of observer performance. Experiments were designed to evaluate the class of methods that correct for scatter by subtracting counts. An anthropomorphic phantom was used with Monte Carlo simulation to simulate liver imaging with labeled antibodies. The lesion was a 2.5-cm-diameter, spherical, “cold” tumor in the liver, a large, “warm” background. Ramp-filtered back-projection and non-iterative Chang attenuation compensation were used to approximate clinical practice. Perfect scatter rejection, defined as images containing only primary (non-scattered) photons, was selected as the ideal case. These images were compared with uncorrected images for conditions of both low and high scatter fractions (SF, the scatter-to-primary ratio), typical of Tc-99m and In-111, respectively. In addition, the dual photopeak window (DPW) method was tested in order to evaluate a non-ideal subtraction correction. Receiver operating characteristic (ROC) experiments were conducted under signal-known-exactly conditions, with the area under the curve, A_z , used as the index of accuracy. A statistically significant difference in detection was found only in a few cases, when scatter rejection was compared with no correction. Subtraction of a scatter estimate changes both contrast and noise, such that an improvement in quantitative accuracy does not necessarily provide improvement in detection accuracy. Corrections that approach scatter rejection conditions may offer some improvement in detection, particularly for cases of high SFs