Parallel List-Mode Reconstruction and Calculation of the System Matrix for the High-Resolution ClearPET Neuro

The list-mode format is particularly suitable for high-resolution tomographic systems, where the precision of the data have not to be compromised, but also claims new demands on data processing and image reconstruction. We apply a ray-tracing algorithm to calculate the three-dimensional coincidence response function. The algorithm is embedded in a list-mode reconstruction framework and has been previously assessed with data from the small-animal ClearPETtrade Neuro scanner. In order to compare and verify our algorithm, exemplary response functions are simulated using GATE. Long reconstruction time has been overcome by implementing a reconstruction framework for distributed computing. The implementation uses the message passing interface for communication and a simple master/slave model with dynamical load balancing of the slaves.