@article {Mou1738, author = {Tian Mou and Jian Huang and Finbarr O{\textquoteright}Sullivan}, title = {Statistical Analysis of the 3-D Imaging Characteristics of a PET Scanner: More Detailed Processing of Routine Quality Assurance Data.}, volume = {59}, number = {supplement 1}, pages = {1738--1738}, year = {2018}, publisher = {Society of Nuclear Medicine}, abstract = {1738Objectives: When a PET scanner is installed and begins to be used operationally, its performance may deviate somewhat from design or installation stage specifications. Thus it is recommended that routine quality assurance (QA) measurements be used to provide an operational understanding of scanning properties. While QA data are primarily used to evaluate sensitivity and bias patterns, there is a possibility to make further use of such data sets in order to achieve a fuller understanding of the 3-D scanning properties. Methods: The techniques are focused on the analysis of data acquired from a uniform cylindrical phantom containing a known activity concentration and scanned over a 15 minute period matched to a clinical FDG protocol using iterative reconstruction. Our approach builds on recent work showing that the distributional characteristics of iteratively reconstructed PET images can be well-described by a Gamma model form capturing the non-Gaussian skewness. The Gamma model allows a probability transformation to be applied to produce normalized values for analysis of covariance patterns. 3-D spatial auto-regressive (SAR) models are used to describe auto-covariance structure. Appropriate likelihood-based statistical techniques for estimation of the autoregressive model coefficients are described. The approach is illustrated on QA data from an operational clinical scanner. A range of simulation studies are used for validation. Results: The analysis approach is readily implemented in R - an open-source statistical computing platform. Data from a 3-D ROI placed within the phantom is used as input. Simulation studies demsonstrate the reliability (consistency) of the analysis scheme as a function of the ROI size and total activity. The behavior of parameter mean square error matches asymptotic theory. Second order SAR modes are found to provide an adequate representation of the autocoveriance characteristics of measured phantom data. The nature of the estimated spatial covariance did not change when a multi-frame dynamic scanning protocol was used. This is in line with expectations. Conclusions: The analysis of uniform phantom measurements, employing a dose-dependent Gamma distribution and simple spatial autoregressive modeling, provides an effective strategy for approximating the 3-D statistical characteritstics of an operational scanner employing iterative reconstruction. The approach leads to a simple mechanism for simulation of scanner data{\textemdash}one that is readily implemented in R. SUPPORT: SFI PI 11/1027; NIH R33CA225310}, issn = {0161-5505}, URL = {https://jnm.snmjournals.org/content/59/supplement_1/1738}, eprint = {https://jnm.snmjournals.org/content}, journal = {Journal of Nuclear Medicine} }