In a recent survey, endocrinologists reported ∼2/3rd of their practice was typically devoted to treatment of diabetes (Healy et al., 2018). Historically speaking, clinical nuclear medicine has devoted a large portion of positron emission tomography (PET) to imaging the radiolabeled glucose analog, [18F]fluorodeoxyglucose (FDG), to measure glucose uptake in a variety of settings. (Delbeke et al., 2006) As an outside observer, one may be forgiven for thinking that both endocrine and PET research are limited by these clinical emphases; however, there is a rich intertwined history of endocrine research using PET imaging. In short, the rapid and continual development of a number of radiopharmaceuticals targeting different receptor, enzyme and small molecule systems has fostered PET imaging of neuroendocrine system actions in vivo in the human brain for several decades.

This review is structured into four main sections to highlight past research using PET to probe neuroendocrine systems in vivo. First, a brief discussion of PET quantification is included with the idea that researchers interested in performing PET imaging can assess rigor of previous work in their field of interest and determine requirements necessary (e.g., PET scanners, cyclotrons, radioligands) to embark on future research. Based on the methodological considerations below, the second and third sections outline PET radioligands that have been developed to measure changes that are regulated by hormone action (e.g., glucose metabolism, cerebral blood flow, dopamine receptors) and actions within endocrine organs or glands such as steroids (e.g., glucocorticoids receptors), hormones (e.g., estrogen, insulin), and enzymes (e.g., aromatase). The final section discusses recent advances in PET scanner technology that will advance our ability to perform novel neuroendocrinology research.

In order to provide an unbiased starting point, the search terms ‘positron emission tomography’ AND ‘brain’ AND ‘endocrine’ were used resulting in 902 references from 1970 to May 1, 2022. Screening and removal of records for review was performed according to the PRIMSA guidelines for systematic review (Figure 1). After removing all duplicates, case studies, review articles and neuroendocrine tumor PET imaging studies that were outside the scope of this review, 321 references remained for retrieval. Neuroendocrine tumor PET imaging is a fertile on-going research area for those interested (Pollard et al., 2020, Bergeret et al., 2019). One could argue that every brain FDG scan is in some manner a neuroendocrine study; however, only PET studies that reported measurement of endocrine hormones in plasma (e.g., insulin, leptin, estrogen, etc.) were included. In a similar manner, the neuroimaging landscape of dopamine and serotonin receptor ligand PET studies is vast and were only included if there was an attempt to correlate PET measures to endocrine hormones or specific endocrine disorders. In addition, for brevity, section two only includes human PET studies. Section three does discuss some preclinical work regarding novel radioligand development. Throughout the review, a note is made for any studies that have an imbalance in recruitment of sex as a biological variable (SABV) (Galea et al., 2020). This is done to highlight potential pitfalls in extrapolating previous studies to the larger populations and areas of research that may benefit from further study in both sexes.

Over the last ∼50 years, developments in imaging technology, novel radioligands and translational research has led to a wealth of knowledge within neuroscience and whole–body physiology. The ability to track radiolabeled proteins that bind to receptors or enzymes in living humans to measure biochemical reactions has profoundly shaped our understanding of both normal physiology and diseases. The advantage of PET imaging over other imaging modalities is that it allows direct quantification of the target of interest using a radiolabeled molecule (radioligand). Depending on what the radioligand is designed to target; receptors, enzymes or a biological process such as glucose metabolism, if appropriate methods are taken, receptor density, enzyme levels or metabolic rates can be quantified. For an in-depth discussion of kinetic modeling of PET radioligands and trade-offs between quantitative methods we refer the reader to the field consensus nomenclature paper and a recent review (Innis et al., 2007, Hooker and Carson, 2019). However, since various outcome measures are reported throughout the review it is worth a brief discussion.

For fully-quantitative PET scanning, a radioligand is injected and arterial sampling is performed to measure the parent (unmetabolized) radioligand and radiolabeled metabolites that will appear during the scan as the radioligand is metabolized by the body. By measuring the fraction of metabolites in blood we can subtract the portion of the arterial plasma curve that contains metabolites, deemed the metabolite–corrected input function. The metabolite-corrected input–function, when combined with an appropriate compartmental model, can account for biological factors such as blood flow and local metabolism, and be used to estimate rate constants between compartments (Figure 2). These rate constants can then be used to estimate the volume of distribution (VT), the ratio of the radioligand in tissue to plasma at equilibrium (Innis et al., 2007, Hooker and Carson, 2019), in whole brain or, if available, structural magnetic resonance imaging (MRI) to segment specific brain regions-of-interest (ROIs). If a 2-tissue compartment model is appropriate for a particular radioligand, the second tissue compartment (C2, Figure 2) is referred to as specifically bound to the receptor or target of interest. The PET image acquired contains both tissue compartments, C1 and C2, and therefore contains non–displaceable signal of free tracer and non-specific binding. If there is a region or organ with no specific binding of the radioligand, this region can be used as a reference region to estimate the non-displaceable binding potential (BPND). Reference regions account for the free tracer and non-specific binding signal present in both the ROI and reference region and eliminate the need for arterial blood sampling. BPND is dependent on the ratio of target receptor or enzyme density, Bmax or Bavail when measuring only available unbound sites, and affinity (Kd) in vivo. Bmax and Kd are important considerations when choosing a radioligand for a particular imaging application. For example, in the brain, extrastriatal dopamine receptors (DR) have a much lower Bmax than striatal DR, therefore a radioligand with a higher affinity (lower Kd) is necessary to increase BPND and signal to noise (SNR) in those extrastriatal regions for accurate quantification.

The historically most widely used technique in clinical nuclear medicine is the radiolabeled glucose analog [18F]fluorodeoxyglucose (FDG). (Delbeke et al., 2006) FDG, once injected, distributes throughout the circulatory system. FDG can then enter different tissues (e.g., brain cells), facilitated by glucose transporters (K1), where it is either returned to the blood pool (k2) or phosphorylated (k3) (Figure 2). Due to the low amount of dephosphorylation in most tissues, FDG in most applications (e.g., glucose metabolism of cancer cells), is considered “trapped” in cells and cells with higher metabolism will have higher “trapping” or uptake of FDG. This trapping feature of FDG is fortuitous for clinical use as it makes imaging and quantification of PET images relatively straightforward by allowing simplification of dynamic imaging analyzed with the 2-tissue compartment model with kinetic rate contestants (K1, k2, k3, and k4) to static imaging quantified with a simple standardized uptake value (SUV) or SUV ratio (SUVR). SUV normalizes uptake in a ROI by injected dose and body weight to facilitate comparisons between individuals, although it does not fully account for differences in arterial input functions between individuals. SUVR, sometimes reported as target-to-background ratio (TBR), often uses a blood compartment to normalize FDG uptake to plasma levels, if the scan is sufficiently late to allow plasma clearance of the radioligand.

For all radioligands, fully quantitative kinetic modeling with arterial blood sampling is the gold standard. When moving to reference region or semiquantitative approaches (e.g., SUV) there may be appropriate bias trade-offs to simplify the PET acquisition procedure as it is desirable to eliminate invasive arterial sampling and the need to have access to a metabolite lab to process blood samples. Typically, these simplified methods are correlated to the gold standard arterial sampling methods to demonstrate similarity in quantitative outcomes prior to subsequent use in prospective studies. Although, many of the studies outlined in this review are primarily research and not clinical diagnostic scans, simplified quantification methods are necessary if a radioligand is going to be used for clinical diagnostics or clinical trials as most PET centers do not perform routine arterial sampling studies. We have described several quantitative methods which assist the reader in assessing rigor of previous work in their field of interest.