Optimization of Maximum Likelihood Estimator Images for PET: II. Statistical Analysis of Human Brain FDG Studies

II. Statistical Analysis of Human Brain FDG Studies J. Llacer, Ph.D), E. Veklerov, Ph.D.l, K.J. Coakley, Ph.D.2, E.J. Hoffman, Ph.D.3 and J. Nunez, Ph.D.4 I Engineering Division, Lawrence Berkeley Laboratory University of California, Berkeley, CA 94720 2Statistical Engineering Division National Institute of Standards and Technology, 3Dept. of Radiological Sciences, School of Medicine, University of California, Los Angeles, and 4Facultat de Fisica, Universitat de Barcelona, Spain The work presented in this paper studies the question of what are the characteristics of Maximum Likelihood Estimator (MLE) reconstructions in Positron Emission Tomography (PET) that make that method superior to the standard Filtered Backprojection (FBP). PET data of human brain Fluorodeoxiglucose studies have been used to evaluate the comparative statistical characteristics of regional bias and expected error of the two methods. The results are that properly reconstructed MLE images are as unbiased as those obtained by FBP and that the expected error in the MLE case is approximately proportional to the square root of the number of counts in a region. In contrast, FBP reconstructions show an expected error that is high and nearly independent of the number of counts in a region. A preliminary study shows that those statistical characteristics of MLE reconstructions translate into improved lesion detectability in regions of low activity for the case of a simple detectability task carried out by non-medical observers.