Modelling and Simulations of a clinical PET-system using the GATE Monte Carlo software

Sammanfattning: Introduction: The increased importance of PET imaging in the health care system has lead to a constant need for improvement and optimisation of scanner systems and their protocols. By utilising the features of the GATE Monte Carlo program one can gain a better understanding of the limitations and possibilities of the system. GATE together with the reconstruction software CASToR provides a link from simulation of modeled system to reconstructed images. The aim of this thesis was to learn the basics of the Monte Carlo method and how to operate GATE, to create a model of a generic PET camera as a first approximation, perform simulations using 18F as radioactive source and from simulated data reconstruct images using the CASToR software. Material and method: The modelling of the generic camera used the G.E. PET/CT Discovery-690 as a template. The NEMA 2012/IEC 2008 PET phantom was used to perform a measurement on a G.E. PET/CT Discovery-690. CT-images from the measurement were segmented and used as a phantom for simulations. Parameters such as true-fraction, scatter-fraction, random-fraction, sensitivity and FWHM of three spheres were compared between the measurement and the simulation. Simulations of a human-like phantom (XCAT) were also performed and compared to images of clinical exams with 18F-fluoride taken with a Philips Gemini TF PET/CT. Comparisons were also made with a SIMIND SPECT simulation of the same phantom in order to study qualitative differences between PET and SPECT. Results: The general structure of the simulated image corresponds well with structures in the images taken by the Philips PET/CT Gemini TF camera. The visual comparison between GATE and SIMIND simulated images show similar geometrical structures. True-fraction, scatter-fraction, random-fraction, sensitivity and FWHM between the measurement, NEMA simulation 1 (energy window threshold 300 keV, crystal energy resolution 20%) and NEMA simulations 2 (energy window threshold 425 keV, crystal energy resolution 8%) (M/S1/S2) were TF: 0.64/0.29/0.47, SF: 0.23/0.45/0.32, RF: 0.14/0.27/0.22, S: 5.1/8.3/4.0 cps/kBq, FWHM for three of the sphere 36.0/36.4/35.9 mm, 27.5/25.4/24.2 mm and 16.0/14.5/14.7 mm. NEMA simulation 1 showed high values of scatter- and random-fraction and a large deviation in sensitivity with 63% higher compared to the measurement. After increasing the energy window threshold to 425 keV and crystal energy resolution to 8% (NEMA simulation 2) the results from the measurements and simulations became more in accordance with each other, but are still deviating. Conclusion: GATE is a powerful tool simulating emission tomography. With the modelling tools provided by the Geant4 kernel, GATE offers the user good possibilities to model specific scanner geometry and set-up. While GATE provides good simulating accuracy it comes at a cost of computational power and time - the need for decreasing simulation time is thus necessary. The link between CASToR and GATE creates a good chain from simulated data to reconstructed images. PET images show better spatial resolution over SPECT images of identical distribution. The results are satisfying with regards to visual properties, however, the quantifiable parameters have a significant deviation that needs to be further investigated as to why the deviations occur. By further improving the model GATE simulations could possibly act as an option in patient studies in the future.