Bayesian Image Reconstruction for the Clear-PEM scanner

Abstract

The Clear-PEM scanner, a device based on a two-head planar detector, is currently being developed for positron emission mammography under the framework of the Crystal Clear Collaboration. Previously, we reported the introduction of new features into the STIR library, allowing its use with Clear-PEM simulated data. Previous reports indicate that image reconstruction of cylindrical PET data using Bayesian reconstruction methods and the median root prior tend to yield smoother images than non Bayesian approaches. To our knowledge, these algorithms have not yet been applied to scanners with non-conventional geometries. We now report some exploratory work regarding the use of Bayesian image reconstruction methods and median root prior with Clear-PEM simulated data. The results presented indicate that the use of the maximum a posteriori algorithm and the median root prior lead to a clear reduction in the noise of the images, without a significant loss of image spatial resolution or decrease in contrast.