Performance Characterization, Image Processing, and Multimodality Coregistration of Small Animal Positron Emission Tomography Systems

Positron Emission Tomography (PET) systems designed specifically for small animal imaging have a unique set of requirements for successful operation. This work investigates a limited angle tomography microPET system, in contrast to conventional cylindrical systems. Initial studies included measurements of spatial resolution, uniformity, contrast-to-noise ratios, resolution recovery coefficients, and detector deadtime.A subsequent investigation was performed to ascertain the use of Factor Analysis in mitigating artifacts caused by the limited angle tomography. Three separate studies were conducted: a dual-isotope experiment, a semi quantitative study, and mouse models. The results indicate potential use of Factor Analysis with possible limitations due to noise or uptake.A more rigorous investigation into the limits of Factor Analysis was performed using Principal Component Analysis. Using the definitions of correlation and covariance matrices, the results of uniform phantom studies were used to measure uniformity, signal-to-noise ratio, and noise distribution. This analysis was applied to the limited angle tomographic system and a conventional cylindrical system. The results were insightful and in strong agreement with Poisson statistics.Multimodality coregistration between systems was also performed using a rigid body least-squares transformation ("Horn's method"). Spherical phantoms were imaged to obtain calibration data within each system. In addition to deriving the transformation between a pair of systems, a metric was computed to estimate the error associated with the coregistration. Once computed, the transformation was applied to independently acquired mice studies from each pair of systems. The method proved fairly successful, and a number of improvements are suggested.The rigid body nature of the transformation also allowed for a convenient method of measuring spatial linearity. Coregistration between synthetic grid-like images, and actual images of grid-like phantoms, were performed. The least-squares metric produced is used to determine spatial linearity. Grid-like phantoms were created using activity-infused ink in conjunction with a conventional inkjet printer. Each phantom was scanned on the limited angle tomography system. The results indicated a very high degree of spatial linearity.Although motivated by limited angle PET, most of the methods employed may be used for other imaging modalities as well. These methods are statistical in nature, and may be similarly interpreted within various modalities. While the purpose of each study was different, the underlying mathematical concepts are common amongst most of them