A direct image reconstruction algorithm for PET scanners based on monolithic crystals

In this work, a direct image reconstruction algorithm for PET is proposed. This method follows naturally from the concept of LOR, and easily admits depth of interaction and time of flight information. In addition, this algorithm does not require the computation of sinograms or a system matrix. Thus, it is not restricted in the formation of high pixel density images and allows the image formation in real-time (simultaneously to the data acquisition process). One of the key advantages of this algorithm is the preservation of high frequencies, producing accurate images that can be generated in high resolution, because of the lack of space restrictions imposed by the intermediate generation of sinograms, aside of the native support of real-time image reconstruction. In this work, a C++ implementation of the described algorithm has been developed. Image reconstructions of simulated PET events according to the MindView (a brain dedicated PET insert) scanner geometry and real data measured with the first one ring prototype system have been performed. Results show capability of correct image formation, either with simulated and real data.