Truncation artifact correction of attenuation map with iterative and model based reconstruction

Attenuation compensation is an important issue for SPECT imaging. For this purpose transmission measurements can be acquired on a gamma, but the acquisitions are usually truncated. The truncation results in large artifacts in the attenuation map when reconstructed with usual filtered-backprojection algorithms. These artifacts prevents good attenuation correction of emission map. The authors propose a new efficient method for truncation correction which is based on an iterative reconstruction/reprojection algorithm using extended projections and constraints on the reconstructed map. Extended projections are a combination of truncated and reprojected projections with a continuity constraint at the truncation edge. The constraints on the attenuation map are based on object model. The first one is a large elliptical model, with positivity constraints. During the iterative reconstruction process, the model is refined to a have geometric shape closer to the patient's outline. The refining is an active contour segmentation technique. Additional constraints on the attenuation factors can finally be incorporated. Results on truncation correction obtained for simulated and experimental data are presented and the effect on attenuation compensation in emission reconstruction is discussed.