A Method for System Matrix Construction and Processing for Reconstruction of In-Beam PET Data

At present, positron emission tomography (PET) is the only available technique for an in-situ, non-invasive monitoring of the dose delivery precision in highly conformal ion beam therapy. At the heavy ion therapy facility at the Gesellschaft fur Schwerionenforschung (GSI) Darmstadt, Germany, an in-beam PET scanner is operated for quality assurance monitoring, simultaneously to the therapeutic irradiation. The dedicated reconstruction algorithm and data acquisition system were developed to fit specific conditions of in-beam PET (dual-head, limited-angle geometry, very low counting statistics). In this work we propose a precise technique for calculating and processing the geometric component of the system matrix and analyze the influence of the system matrix on the quality of reconstructed in-beam PET image. We show that for in-beam PET data the maximum likelihood expectation maximization (MLEM) algorithm is robust to errors in the system matrix and produces acceptable results even for a system matrix calculated with poor quality, whereas the quality of images reconstructed with the randomly filled subsets expectation maximization algorithm (RFS-EM) depends more on the accuracy of the system matrix.