Reconstruction of 3-D Positron Emission with Maximum Likelihood

摘要

The statistical principle of maximum likelihood (ML) has recently been applied to positron image reconstruction with devices that obtain a two-dimensional (2-D) image, (e.g., a single cross-section of a human head). We have explored implementation of ML in simulated data from a three-dimensional (3-D) device consisting of planar parallel gamma cameras. A coarse image of 3-D rectangular modules simulated with a perfect detector had good recovery of edges and average count intensity even with count rates under 20 emissions per individual volume element. Using the summary statistics from that exploration, a complex 3-D object that corresponds to an experiment in human physiology was investigated. Recovery of that image depends strongly on details of camera performance included in the simulation and ML algorithm. The best algorithm depended on which image features were desired from the data. Quantitative recovery of actual complex images having less than 100 emissions per element is not feasible with current techniques. Keywords: Image; Positron; Tomography; Maximum likelihood; Planar detectors. (jhd)