Choosing the ART relaxation parameter for Clear-PEM 2D image reconstruction

Abstract

The Algebraic Reconstruction Technique (ART) is an iterative image reconstruction algorithm. During the development of the Clear-PEM device, a PET scanner designed for the evaluation of breast cancer, multiple tests were done in order to optimise the reconstruction process. The comparison between ART, MLEM and OSEM indicates that ART can perform faster and with better image quality than the other, most common algorithms. It is claimed in this paper that if ART's relaxation parameter is carefully adjusted to the reconstruction procedure it can produce high quality images in short computational time. This is confirmed by showing that with the relaxation parameter evolving as a logarithmic function, ART can match in terms of image quality and overcome in terms of computational time the performance of MLEM and OSEM algorithms. However, this study was performed only with simulated data and the level of noise with real data may be different.