MLE reconstruction of a brain phantom using a Monte Carlo transition matrix and a statistical stopping rule

Publisher Website
Google Scholar
Abstract
In order to study properties of the maximum-likelihood-estimator (MLE) algorithm for image reconstruction in positron-emission tomography (PET), the algorithm is applied to data obtained by the ECAT-III tomograph from a brain phantom. The procedure for subtracting accidental coincidences from the data stream generated by this physical phantom is such that the resultant data are not Poisson distributed. This makes the present investigation different from other investigations based on computer-simulated phantoms. It is shown that the MLE algorithm is robust enough to yield comparatively good images, especially when the phantom is in the periphery of the field of view, even though the underlying assumption of the algorithm is violated. A stopping rule derived earlier and allowing the user to stop the iterative process before the images begin to deteriorate is tested. Since the rule is based on the Poisson assumption, it does not work well with the presently available data, although it is successful with computer-simulated Poisson data.