Analysis of Attenuation-Correction Efficacy for Tumor Detection in SPECT

A 2004 paper by Bai and Shao offered predictions about the relative effectiveness of attenuation correction (AC) for SPECT oncology that were based on an analysis of tumor-to-background ratio (TBR) for filtered backprojection (FBP). The purpose of our work was to examine the validity of their conclusions on the basis of human-observer performance in localization ROC (LROC) studies. Our methodology involves the calculation of localization probabilities that are conditional on tumor location. Given a particular reconstruction strategy, these probabilities permit assessments of how observer performance is influenced by image properties such as TBR, signal-to-noise ratio, structural noise, and the dynamic range of the image pixels. Simulated mediastinal studies were considered for this paper. Our results corroborate conclusions from Bai and Shao that accurate AC leads to improved tumor detection in the mediastinum, showing this to be true nearly everywhere within that region of the torso.