Date of Degree
PhD (Doctor of Philosophy)
John J. Sunderland
Reconstructed tomographic image resolution of small animal PET imaging systems is improving with advances in radiation detector development. However the trend towards higher resolution systems has come with an increase in price and system complexity. Recent developments in the area of solid-state photomultiplication devices like silicon photomultiplier arrays (SPMA) are creating opportunities for new high performance tools for PET scanner design.
Imaging of excised small animal organs and tissues has been used as part of post-mortem studies in order to gain detailed, high-resolution anatomical information on sacrificed animals. However, this kind of ex-vivo specimen imaging has largely been limited to ultra-high resolution μCT. The inherent limitations to PET resolution have, to date, excluded PET imaging from these ex-vivo imaging studies.
In this work, we leverage the diminishing physical size of current generation SPMA designs to create a very small, simple, and high-resolution prototype detector system targeting ex-vivo tomographic imaging of small animal organs and tissues.
We investigate sensitivity, spatial resolution, and the reconstructed image quality of a prototype small animal PET scanner designed specifically for imaging of excised murine tissue and organs. We aim to demonstrate that a cost-effective silicon photomultiplier (SiPM) array based design with thin crystals (2 mm) to minimize depth of interaction errors might be able to achieve sub-millimeter resolution. We hypothesize that the substantial decrease in sensitivity associated with the thin crystals can be compensated for with increased solid angle detection, longer acquisitions, higher activity and wider acceptance energy windows (due to minimal scatter from excised organs).
The constructed system has a functional field of view (FoV) of 40 mm diameter, which is adequate for most small animal specimen studies. We perform both analytical (3D-FBP) and iterative (ML-EM) methods in order to reconstruct tomographic images. Results demonstrate good agreement between the simulation and the prototype. Our detector system with pixelated crystals is able to separate small objects as close as 1.25 mm apart, whereas spatial resolution converges to the theoretical limit of 1.6 mm (half the size of the smallest detecting element), which is to comparable to the spatial resolution of the existing commercial small animal PET systems. Better system spatial resolution is achievable with new generation SiPM detector boards with 1 mm x 1 mm cell dimensions.
We demonstrate through Monte Carlo simulations that it is possible to achieve sub-millimeter spatial image resolution (0.7 mm for our scanner) in complex objects using monolithic crystals and exploiting the light-sharing mechanism among the neighboring detector cells. Results also suggest that scanner (or object) rotation minimizes artifacts arising from poor angular sampling, which is even more significant in smaller PET designs as the gaps between the sensitive regions of the detector have a more exaggerated effect on the overall reconstructed image quality when the design is more compact. Sensitivity of the system, on the other hand, can be doubled by adding two additional detector heads resulting in a, fully closed, 4π geometry.
In this study, we design, simulate, and construct a novel, compact, high resolution, and cost-effective PET imaging device designed specifically for imaging of excised murine tissue and organs. The prototype scanner is a four-face parallel-plate PET scanner based on silicon photomultiplier (SiPM) detectors coupled with 2 mm thick LYSO crystals. We implement a GATE/Geant4 Monte Carlo simulation of this scanner to validate and analyze its properties using several crystal configurations including individual crystals directly coupled to SiPM array pixels and monolithic crystals covering the SiPM arrays. Based upon simulation, we identify a central 40mm diameter region as the scanner's field of view (FoV), where resolution is only minimally impacted by parallax. The scanner system with pixelated LYSO crystals is able to separate small objects as close as 1.25 mm around the center and 1.50 mm at the periphery of the defined FoV. Simulations demonstrate the necessity of detector (or specimen) rotation in order to remove the artifacts resulting from poor angular sampling and achieve acceptable image quality. When studying the monolithic crystal configuration, we first measure precise crystal-photon interaction positioning information from collimated beam experiments with the monolithic crystals using the scintillated light energy sharing mechanism among the neighboring detector pixels. We achieve approximately 0.5 mm positioning accuracy within the monolithic crystal using a regression model over energy distributions. This method allows us to reach a sub-millimeter spatial resolution of 0:7 mm using monolithic crystals.
publicabstract, High Resolution PET, Monte Carlo Simulation, PET, Positron Emission Tomography, SAI, Small Animal Imaging
Copyright 2016 Levent Sensoy