Implementation and performance evaluation of iterative reconstruction algorithms in SPECT: A simulation study using EGS4

Abstract

Three iterative reconstruction algorithms (Maximum Likelihood-Expectation Maximization (MLEM) algorithm, Multiplicative Simultaneous Iterative Reconstruction Technique (MSIRT) and Additive Simultaneous Iterative Reconstruction Technique (ASIRT)) were implemented incorporating Attenuation Correction (AC) and Scatter Correction (SC) for Single Photon Emission Computed Tomography (SPECT). The purpose of this study is to investigate the convergence properties of the three iterative methods by computer simulation using EGS4. Digital brain and cylindrical phantoms were designed for activity distribution and linear attenuation coefficient maps. In this study, the simulated radioisotope was Tc-99m (photon energy = 140 keV). The photon attenuation and scatter and the distance-dependent blurring due to the collimator were involved in the simulated SPECT system. In the implemented reconstruction algorithms, modeling of the effect of attenuation and scattering were involved but the distance-depend blurring was not included. The progress of the reconstruction was monitored by observing the residual squares error (Chi(exp 2)) between the calculated projections and measurement data at each iteration. The number of iterations needed to reach a constant Chi(exp 2) value were different for each algorithm. ASIRT needed the largest number, MLEM the lowest number of iteration. The ASIRT required more than 100 iteration to produce a comparable result of MLEM and MSIRT. Almost equivalent performance was observed between MLEM and MSIRT. All iterative reconstruction algorithms were effective in compensating for non-uniform attenuation distribution. There is potential in the iterative reconstruction algorithms for improvement of quantitative SPECT capability. EGS4 is powerful tool for investigating the properties of reconstruction algorithms without actual phantom experiments using the radioactive isotopes.