• PET
• Monte Carlo
• tomographic image reconstruction

Positron Emission Tomography (PET) is a metabolic imaging modality which measures the distribution of a radiotracer (a compound labeled with a positron emitting radionuclide) in a living subject by the detection of gamma-rays, which are emitted due to the positron-electron annihilations. Depending on the biological and chemical characteristics of the compound, various functional processes within the living subject can be studied. Apart from clinical PET, as a "routine imaging modality" in nuclear medicine, small-animal PET has become an important tool for preclinical studies, e.g. for the evaluation of new radiotracers and therapies or to study receptor bindings, gene expressions, ecc.
However, sophisticated PET scanners are necessary to scan small animals, such as a 30-g mouse, appropriately. The main requirements are a high spatial resolution, to resolve small structures in the reconstructed tracer distribution and a high sensitivity, to discriminate small changes and to be able to detect low doses of the radiotracer.

An integral part of all PET scanners is a reconstruction algorithm, which reconstructs the three dimensional radiotracer distribution from the measured gamma-rays. One purpose of this thesis was the implementation of such a reconstruction algorithm for the YAP-(S)PET scanner. Since an iterative reconstruction algorithm was implemented (maximum likelihood - expectation maximisation), a focus was set on setting up the system model, which is used for the forward- and backprojection during the reconstruction. The model is calculated efficiently and accurately with the newly formulated multi-ray integration method and the results are compared to other integration methods such as Monte Carlo.

Moreover, as the pixelated scintillation crystals of the YAP-(S)PET detectors limit the performance of the scanner, a new design is proposed which is based on slabs of continuous scintillation crystals which are coupled to silicon photomultipliers. First, a single detector head is optimised to improve the intrinsic detector spatial resolution. Afterwards, hit estimation methods are derived to estimate the gamma-ray interaction within the continuous crystal. Finally, the performance of the new design is studied by reconstructing coincidence events from a point source which is simulated with a Geant4 based Monte Carlo simulation.