Towards a stopping rule for the OSEM algorithm for iterative image reconstruction in PET

We have studied certain properties of the OSEM algorithm for iterative image reconstruction in PET. This work is within the framework of our principal objective that is the development and evaluation of a stopping criterion for this iterative algorithm. To investigate the problem of the deterioration of the image quality after several iterations, the statistical properties of the pixel updating coefficients have been studied. Preliminary results show that the values of these coefficients for the pixels that correspond to areas in the source with non-zero activity distribution form a histogram composed of a peak region around to 1.0 and a tail below this value. Given the different rate of convergence of the pixel values within an image, this tail distribution corresponds to those pixels that are yet far from reaching their corresponding true values. We demonstrate that the minimum value of the tail distribution is directly related to the quality of the reconstructed images. The possibility of exploring this observation in order to formulate a stopping criterion for the OSEM algorithm is discussed.