Segmented attenuation map reconstruction from incomplete transmission data

Accurate reconstruction of attenuation maps from incomplete sinograms are required in some cases of SPECT and PET imaging. This paper proposes a new method to directly reconstruct segmented attenuation maps from the incomplete sinograms. The proposed method is based on an image labeling technique where an optimum label configuration minimizing some energy function is found by using a stochastic sampling with simulated annealing. Unlike ordinary image labeling techniques, however, the authors introduce a powerful constraint called the topology-preserving constraint. This constraint reduces the search space to a set of label configurations having the same topology as the known topology of the attenuation map in question. The experimental results demonstrate that the topology-preserving constraint is powerful enough to reconstruct accurate segmented attenuation maps from the incomplete sinograms. The proposed method is tested with simulated data and PET transmission data.