In-beam PET Performance Simulation And Image Reconstruction

Due to reversal-range energy loss curve and high relative biological effect, heavy ions caneffectively achieve accurate cancer treatment. In-Beam positron emission tomography (in-beamPET) is currently the only method for an in-situ monitoring of highly tumor-conformed chargedhadron therapy. In radiotherapy, the clinical effect of deviations from treatment planning is highlyminimized by implementing safety margins around the tumor and selecting proper beam portals.Nevertheless, in-beamPETisabletodetecteventual, undesirablerangedeviationsandanatomicalmodifications during fractionated irradiation, to verify the accuracy of the beam portal deliveredand to provide the radiotherapist with an estimation of the difference in dosage if the treatmentdelivered differs from the planned one.Seventies of last century, in-beam PET research in Berkeley was abandoned due to detectoractivation(BGO),ninetiesGSIusingX,Yusedscanningmagnettocontrolbeam,doublesphericalcap type in-beam PET was successfully implemented in GSI to online monitor cancer therapy.Institute of Modern Physics, Chinese Academy of Sciences, successfully carried out shallowand deep tumor treatment using the high-energy heavy ion provided by HIRFL-CSR. In cope withheavy ion cancer therapy, we carried out a study of dual head in-beam PET. Meanwhile, a set ofsimulation and fully-3D reconstruction routines were developed to test the ability of in-beam PETto monitor the radiation dose distribution during treatment.The in-beam PET based a state-of-the-art scintillator (LYSO), which has fast decay time(40ns) and high light yield (27000ph/MeV), and8×8multi-anody position sensitive detectors(H8500) were built. The in-beam PET contains two plate heads, which are made of3×5detectormodules, respectively. Each detector module is consisted of a22×22array of2×2×15mm3LYSOpixels coupled to a Hamamatsu H8500PMT. The two plate heads are operating in coincidencemode, resulting in72602lines of response within the field of view. In order to readout the twodetectors operated in coincidence, either in standalone mode or at the IMP deep terminal, a multi-channel, zero-suppressing free, list mode data acquisition system was built.The objectives pursued with the present detector geometry are the optimization of in-beam PETappliedforthemonitoringofheavyiontumorirradiation. Thisoptimizationrequires,inafirststep, identifyingpresentorfuturelimitationsofin-beamPET.Presentlimitationsarethosealreadyexisting at the in-beam PET installed at the IMP facility. Future limitations are those expected toarise at in-beam PET to be installed onto heavy ion tumor treatment facilities under planning orconstruction elsewhere. In a second step, the understanding of the sources of those limitations isnecessary. Based on this knowledge several mathematical and technological innovative solutionsare proposed, or extended from its present status, constructed and verified, either by simulation orexperiment.The most important limitation of in-beam PET, firstly addressed in this work, is the problemof image artifacts arising from limited angle tomography. In order to study the origin of severalartifactsandtobeabletoproposeadetectorconfigurationthatminimizesthem, twosoftwaretoolswere need. These were a tomography simulation, capable of handling several detector configura-tions, and a flexible reconstruction routine able of reading and treating mathematically the outputof this simulation. Reconstruction routines provided by conventional PET applied to data sets ofvery low statistics yield images of poor quality. More importantly, they require the tomographyto be space invariant, a requisite only met by closed ring detector geometries. On the other hand,the reconstruction routine developed for the in-beam tomography at IMP is based on a fixed his-togram size. The size of this histogram is given by the product of the number of detectors on bothheads. The histogram size increases almost two orders of magnitude, when using larger heads. Afixed histogram strategy, if implemented, would be too demanding in terms of computer mem-ory and processing time. Therefore, the development of a reconstruction routine that extends themathematical data treatment already implemented for in-beam tomography at IMP was necessary.This extended reconstruction routine, applied onto larger tomographs, allows to study the sourceof artifacts in limited angle tomography.