Reducing the computational load of iterative SPECT reconstruction

Abstract

By accurately modeling the physics of photon transport into the projection and backprojection steps, iterative SPECT reconstruction methods can reduce the degrading effects of attenuation and the distance-dependent detector response. Unfortunately, this greatly increases computation time. Here, the authors describe two approaches which allow SPECT iterative reconstruction algorithms to be implemented with a reduction in the number of computations needed. The first approach requires a preprocessing step prior to iterative reconstruction in which a frequency distance principle (FDP) filter is applied to transform the projection data to those that would have been obtained with a stationary blurring function. The second approach uses a projector backprojector pair which efficiently models distance-dependent blurring using the FDP. Simulation studies show that the FDP pre-processing approach exhibits the useful property of uniform convergence across the reconstructed image. However, at higher iteration numbers undesirable noise correlations become evident. The approach based on using an FDP based projector backprojector pair gave reconstructions which visually appeared to have similar noise characteristics and a slightly higher rate of convergence compared to reconstructions obtained with the standard brute-force approach of modeling the detector response into the projection and backprojection operations. Both reconstruction approaches can be implemented with little increase in computation time over that needed for iterative reconstruction without modeling detector response and attenuation. >