The simulation of concurrent iterative reconstruction algorithms for single photon emission computed tomography

Iterative reconstruction algorithms for emission computed tomography potentially offer significantly improved qualitative and quantitative images when compared to those generated by the analytic methods commonly used today, however, iterative techniques require extensive computational resources for use in clinical environments. The authors have developed a new class of iterative reconstruction algorithms designed such that they can be executed concurrently with the acquisition of emission projection data. This feature can dramatically reduce the number of post-acquisition iterations required to reconstruct a complete set of projection data. An important aspect of these algorithms is the method used to select which projections are processed during each concurrent iteration. The method used to partition projections into subsets impacts, among other issues, the convergence rate, stability and spatial resolution of the concurrent algorithms. Via simulations of concurrent SPECT reconstruction, the authors have demonstrated the feasibility of the technique using two projection subset construction methods (independent and sequential subsets). Their preliminary results indicate that both subset construction methods can be used to shorten the post-acquisition reconstruction time, although sequential subsets are generally more efficient. The authors also observe that if SPECT projections are acquired over a period of T seconds, the post-acquisition reconstruction time can be reduced by at least (1/3)T seconds