GPU-based acceleration of the MLEM algorithm for SPECT parallel imaging with attenuation correction and compensation for detector response

Abstract Parallel projection based Single Photon Emission Computed Tomography (SPECT) is one of the most widely used nuclear imaging technique even nowadays. Serious artefacts are produced in the reconstructed images due to the non-homogeneous attenuation medium and the distance dependent spatial resolution (DDSR) of the parallel imaging. Effective non-uniform attenuation correction and DDSR reduction procedures should be applied in order to improve the SPECT image quality. We have developed a novel parallel reconstruction method using the Maximum Likelihood Expectation Maximization iterative reconstruction algorithm with attenuation correction and compensation for the DDSR effect in the forward projector. In order to compensate the well-known extreme computation intensity of this reconstruction method a parallel version of the algorithm is created where the computation tasks of the algorithm are executed simultaneously on a GPU. By this reduction of the running time this accurate reconstruction algorithm become available for the use in the clinical applications. The algorithm has been verified using simulation studies.