Data detected by a positron emission tomography (PET) scanner is translated into concentration of radiotracer by using image reconstruction methods. Applying reconstruction algorithms will only retrieve the radiotracer concentration. However, there will be no discrimination between signals arising from specific binding, non­specific binding, free tracer or tracer in plasma. Compartmental models, expressed by systems of differential equations, place the focus on the isolation of the signal of interest. Various such models and related graphical analysis methods have been proposed, with different model and experimental designs as well as varying degrees of invasiveness. As a common trait, all these quantification methods require added information, notably an arterial plasma input function (AIF). The gold standard in PET quantification is using cannulation to draw arterial blood and determine the AIF. Unfortunately, this is not ideal practice as it is sensitive to errors, has a risk of adverse effects and complicates the logistics. Also, it requires time­consuming metabolite measurements and there is a risk of radiation exposure for technical staff. As such, arterial cannulation is an uncomfortable and invasive procedure, discouraging patients and volunteers from participating in PET scans.

Image derived input function (IDIF) methods are an attractive non­invasive alternative to arterial blood sampling. In brain PET imaging, the idea of IDIF methods is to acquire the AIF from the dynamic PET data by looking into intra­cranial blood vessels such as the internal carotids. In the literature, the method used to compute an IDIF from the internal carotid arteries for brain PET studies requires three steps: 1) carotid identification, 2) whole­blood time activity curve estimation from the carotid identified area and 3) correction for plasma to whole­blood ratio and for metabolites. Unfortunately, several limitations hinder the applicability of this approach for some radiotracers.

We focus on the radiotracer 11C­Raclopride, for which no IDIF methods had been developed yet, but validated reference region­based compartmental models do provide a reference solution. We propose and assess non­invasive methods to allow for the accurate quantification of PET images without needing to resort to arterial sampling. The results are derived from a computational phantom and from ongoing patient studies.