Reconstruction, quantification and standardization methods in positron emission tomography

Resumo

Although the information from the PET technique can be exploited both qualitatively and quantitatively, quantitative analysis provides reliable and reproducible numerical measures that can not be obtained qualitatively. Nevertheless, significant efforts need to be made to ensure fully correct information, due to the fact that both image quality and quantitative accuracy of PET can be degraded by several physical factors. The general purpose of the current thesis was the development of processing, quantification and standardization methods, aimed at enhancing the quality and accuracy of Positron Emission Tomography (PET) images, both for human brain studies and for high-resolution small animal studies. A study of the effect of the degradations on human PET studies showed that the presence of the scatter coincidences causes a subestimation of 12% on the quantification of relative uptake changes in 18F-FDG brain studies. Furthermore, the effect of scatter, anatomical variability and the fully 3D reconstruction in the statistical parametric mapping output of 18F-FDG neuroactivation PET studies was also investigated. A scatter correction and fully-3D reconstruction improved the detection sensitivity. Thus, the sample size was reducing up to 18% when the scatter correction was performed and between 8-33% when fylly-3D reconstruction was carried out. The effect of several degradations on the quantification of 18F-DOPA dynamic PET studies was also investigated and our results showed that scatter and partial volume effect are essential corrections to reach an accurate quantification. A scatter correction method based on the Single Scatter Simulation algorithm was implemented it is currently included into analytic and iterative reconstruction algorithms of the STIR library and it is available to the PET community under the Lesser FNU Public License. Furthermore, a Double Scatter Simulation algorithm was also implemented and our results showed the total scatter distribution can be obtained from the double scatter simulation by scaling to the total events when large objects are scanned. Due to the spatial resolution requeriments for high-resolution small animal PET scanners the suitability of SimSET and GATE Monte Carlo codes for simulating these studies was also assessed. Our results showed that SimSET can be used to simulate the detector response with medium accuracy and they are in agreement whit GATE simulations in terms of the scatter fraction (5.46% versus 5.54%).Furthermore, a new fully 3D iterative reconstruction algorithm which achieved a significant underestimation of the activity values in the body with respect to the head and the scatter correction is not an essential correction for the small animal PET studies and only in heart studies in rats could be needed.