Date of Degree
PhD (Doctor of Philosophy)
Mark T. Madsen
Positron emission tomography (PET) and single photon emission tomography (SPECT) are two nuclear emission-imaging modalities that rely on the detection of high-energy photons emitted from radiotracers administered to the subject. The majority of these photons are attenuated (absorbed or scattered) in the body, resulting in count losses or deviations from true detection, which in turn degrades the accuracy of images. In clinical emission tomography, sophisticated correction methods are often required employing additional x-ray CT or radionuclide transmission scans. Having proven their potential in both clinical and research areas, both PET and SPECT are being adapted for small animal imaging. However, despite the growing interest in small animal emission tomography, little scientific information exists about the accuracy of these correction methods on smaller size objects, and what level of correction is required.
The purpose of this work is to determine the role of attenuation and scatter corrections as a function of object size through simulations. The simulations were performed using Interactive Data Language (IDL) and a Monte Carlo based package, Geant4 application for emission tomography (GATE). In IDL simulations, PET and SPECT data acquisition were modeled in the presence of attenuation. A mathematical emission and attenuation phantom approximating a thorax slice and slices from real PET/CT data were scaled to 5 different sizes (i.e., human, dog, rabbit, rat and mouse). The simulated emission data collected from these objects were reconstructed. The reconstructed images, with and without attenuation correction, were compared to the ideal (i.e., non-attenuated) reconstruction. Next, using GATE, scatter fraction values (the ratio of the scatter counts to the total counts) of PET and SPECT scanners were measured for various sizes of NEMA (cylindrical phantoms representing small animals and human), MOBY (realistic mouse/rat model) and XCAT (realistic human model) digital phantoms. In addition, PET projection files for different sizes of MOBY phantoms were reconstructed in 6 different conditions including attenuation and scatter corrections. Selected regions were analyzed for these different reconstruction conditions and object sizes. Finally, real mouse data from the real version of the same small animal PET scanner we modeled in our simulations were analyzed for similar reconstruction conditions.
Both our IDL and GATE simulations showed that, for small animal PET and SPECT, even the smallest size objects (~2 cm diameter) showed ~15% error when both attenuation and scatter were not corrected. However, a simple attenuation correction using a uniform attenuation map and object boundary obtained from emission data significantly reduces this error (~1% for smallest size and ~6% for largest size, in non-lung regions). In addition, we did not observe any significant improvement between the uses of uniform or actual attenuation map (e.g., only ~0.5% for largest size in PET studies). The scatter correction was not significant for smaller size objects, but became increasingly important for larger sizes objects.
These results suggest that for all mouse sizes and most rat sizes, uniform attenuation correction can be performed using emission data only. For smaller sizes up to ~ 4 cm, scatter correction is not required even in lung regions. For larger sizes if accurate quantization needed, additional transmission scan may be required to estimate an accurate attenuation map for both attenuation and scatter corrections.
Copyright 2010 Arda Bekir Konik