Quantitative SPECT brain imaging: effects of attenuation and detector response

Two reconstruction methods that compensate for attenuation and detector response, a 3-D maximum-likelihood (ML) expectation-maximization (EM) method and a filtered backprojection (FB) method with Metz filter and Chang attenuation compensation, were implemented and compared in terms of quantitative accuracy and image noise. The methods were tested on simulated data for the 3D Hoffman brain phantom. The simulation incorporated attenuation and distance-dependent detector response. Bias and standard deviation of reconstructed voxel intensities were measured in the gray- and white-matter regions. The results for ML showed that, in both regions, as the number of iterations increased, bias decreased and standard deviation increased. Similar results were observed for FB as the Metz filter power increased. In both regions, ML had a smaller standard deviation than FB for a given bias. Reconstruction times for the ML method have been greatly reduced through efficient coding and limited source support, and by computing attenuation factors only along rays perpendicular to the detector