Population derived principle component analysis based model for the [<sup>18</sup>F]PBR111 arterial input function in rats

Abstract

Translocator protein (TSPO) PET ligands such as [ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> F]PBR111 with high specificity for TSPO or formerly the peripheral benzodiazepine (PBR) receptor, are frequently used as a brain inflammation biomarker. In this study a population derived model for the metabolite corrected arterial plasma input function (IF) of [ <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup> F]PBR111 in rat is developed. Such population derived models are of interest as they allow to obtain the IF using a limited number of samples and do not require draining the animal for blood. We have therefore developed and evaluated two models based on principle component analysis (PCA) of the population data. The first model follows the conventional approach and builds a model of the IF. The second model, introduced here, builds a model for the area under the curve (AUC) of the IF. The coefficients for this newly developed model are estimated using multivariate linear regression analysis of the population data. The two approaches are evaluated using in vivo data from rats (n=8) comprising both manual sampling and the use of an extracorporeal arterio-venous shunt and coincidence detector system. Optimal time points for the manual sampling of a single or of two samples are estimated. Performance is evaluated in terms of the error on the estimated AUC and the error on the estimated total volume of distribution calculated using the Logan graphical analysis. We find that the proposed PCA model for AUC outperforms the PCA model for IF. This is of interest for cases where one models the reversible tracer binding using the Logan plot that requires AUC(t) rather than IF(t).