· PET and fMRI have many complementary measurements with utility for neuroscience and (potentially) neuropathology

· Function versus occupancy now can be done simultaneously in vivo

· These scanners improve the efficiency for flow-metabolism or oxygen/glucose comparisons

Scientists and clinicians seeking to combine fMRI with the molecular imaging of PET in order to

· Better understand the relationship between our imaging tools and neurobiology

· Develop/employ novel biomarkers for understanding brain function in health and disease

Focusing primarily upon “challenge” studies (task or drug), this talk addresses

· PET capabilities and limitations: temporal/spatial resolution, detection power, ligand availability

· PET analysis models for neurotransmitter mapping, and the “binding potential”

· Strategies for developing paradigms for combined functional PET/fMRI

· What can PET contribute to fMRI?

· What can fMRI contribute to PET?

The recent generation of PET/MR scanners has facilitated two classical paradigms that previously were difficult to accomplish in vivo: 1) function (fMRI) versus receptor occupancy (PET), and 2) flow (fMRI) versus metabolism (PET). The first paradigm provides a unique tool for studying drugs and task-induced endogenous neurotransmitter release. The second paradigm has been addressed in a limited number of older PET studies but now can be widely and efficiently employed using updated techniques. Both paradigms have utility for interpreting neuroscience and clinical investigations. This talk will address recent developments and directions using these approaches.

fMRI has become the tool of choice for mapping changes in brain function due to excellent spatiotemporal resolution and good detection power in cases where repetitive averaging can be employed. For functional brain studies using PET, temporal resolution is more restricting than spatial resolution. On existing PET/MR scanners, PET point-spread functions range from about 1 mm on preclinical systems to 4.5 mm on whole-body clinical systems [1]. Temporal resolution, on the other hand, is limited by both signal-to-noise ratios and radiotracer pharmacokinetics, which set limits of about one minute on individual time points but tens of minutes or more to determine a binding potential.

Using reversible radiotracers that bind to neuroreceptors, PET analyses generally employ compartmental models assuming first-order kinetics to derive a “binding potential”, which is the product of the unbound (or “available”) receptor concentration and the ligand-receptor affinity. This outcome measure conflates three underlying parameters: receptor concentration, basal neurotransmitter level (which masks some fraction of receptors), and the binding strength between ligand and receptor. In principle, each of these parameters can change during a functional challenge and can differ across subject populations. In clinical studies, differences in dopaminergic binding potentials have been attributed to differences in receptor densities [2] or basal neurotransmitter levels [3]; both possibilities must be considered. Changes in ligand-receptor affinity are thought to underlie some of the paradoxical increases in binding potentials observed using agonist challenges with some radiotracers [4, 5].

One of the challenges of designing PET/fMRI for human studies is these modalities operate optimally at very different temporal scales: BOLD fMRI excels using traditional on/off stimulus designs but is poor at detecting slowly-evolving pharmacologically-induced responses, and PET is better suited to long/continuous stimuli like drug challenges and relatively poor at detecting task-induced activity. Preclinical studies using drug challenges can boost fMRI detection power using exogenous contrast agent, and ASL offers a potential alternative to BOLD fMRI for drug studies in humans [6]. Many of the early PET/MR functional studies have employed preclinical non-human primates, but applications in human work are rapidly proliferating.

Initial PET/fMRI studies in our laboratory primarily have employed preclinical NHP models and investigated fMRI and PET signals using antagonists, agonists, and indirect agonists that elevate endogenous neurotransmitter. Using a dopamine D2-receptor antagonist challenge, we observed a monotonic relationship between occupancy and function in basal ganglia that was matched in space and time [7], a result consistent with a classical occupancy model in which binding drives function in the absence of effects like desensitization. Results of that study suggested that the ratio of changes in fMRI signal and PET binding potential might provide an index of basal receptor occupancy, which cannot be easily be measured by PET alone, and we have begun translational studies.

Based upon a classical model informed by PET studies, we developed a multi-receptor model of dopamine-induced fMRI signal that consistently explains a wide range of preclinical data that previously were viewed as inconsistent or divergent [8]. Using a high-affinity agonist, we observed a temporal divergence between PET and fMRI markers of occupancy and function, suggestive of neuro-adaptive mechanisms including desensitization and receptor internalization [9].

Recent PET/fMRI studies suggest the potential to efficiently assay flow-metabolic coupling using combined information. fMRI using calibrated BOLD signal provides a way to estimate relative CMRO2 with reasonable temporal resolution, and a recently developed dynamic FDG-PET method now provides a way to assess glucose metabolism on a similar time scale using multiple within-session stimuli in human subjects [10]. Moreover, the assumption that metabolism is primarily driven by post-synaptic processes [11] has led to a novel combination of BOLD resting-state fMRI with FDG-PET that attempts to determine the directionality of functional connectivity changes [12].

For functional brain studies, simultaneous PET/MR combines the efficient but non-specific mapping tool that is fMRI with the molecular capabilities of PET. Among the many possibilities for leveraging the complementary information from these modalities, this presentation focuses on neurovascular coupling studies that fall within the general bounds of function-occupancy or flow-metabolism associations.

Prior to the advent of these PET/MR scanners, there were virtually no studies of fMRI versus occupancy even though these modalities have existed separately for decades. Occupancy-function studies are fundamental to pharmacodynamics and now can be accomplished in vivo. One way to exploit this new capability is to use occupancy studies as way to interpret fMRI signal changes, as in our multi-receptor model of dopamine-induced fMRI signal in basal ganglia [8]. Alternatively, the functional readout from fMRI can be used to help clarify mechanisms underlying changes in binding potentials. For instance, our initial studies using dopaminergic drugs have addressed a conundrum that has been evident in PET studies for some time: using PET radioligands that bind to neuroreceptors, infusion of antagonists robustly reduce binding potentials, but agonists so often give small or even paradoxical changes in binding potential – why is this? Speculation has centered upon two hypotheses [4, 5]: 1) agonists bind preferentially to high-affinity receptor states, and so antagonist radiotracers are less sensitive to agonist binding, or 2) agonists induce internalization, leading to changes in radioligand-receptor affinity that confound interpretation of changes in binding potential. Having a functional readout in the form of fMRI provides new avenues for exploring these questions.

PET/fMRI offers many avenues for cross-validating measurements techniques (e.g., CMRO2 or CBF), but perhaps the most obvious complementary metabolic measurements to be performed are BOLD-based estimates of relative oxygen utilization together with PET measurements of glucose utilization. Due to the difficulty of estimating CMRO2 measurements by PET, there still is a lack of agreement about the extent of anaerobic glycolysis in the brain. Moreover, FDG-PET, perhaps using the newer dynamic uptake paradigms, might serve as a useful surrogate to fMRI studies generally by helping define both resting-state metabolism and task-induced changes in metabolism.

PET/fMRI combines one very efficient tool for assessment of changes in brain activity (fMRI) with a molecular imaging modality that is less efficient but offers very targeted neurochemical probes. For fMRI, PET offers the ability to help clarify fMRI signal mechanisms associated with various neurotransmitter systems (e.g., dopamine, serotonin, opioid) or with tasks that elevate endogenous neurotransmitter. The anatomical and functional information provided by MRI benefits in PET in a multitude of ways. These still are early days for PET/fMRI, but expect a rapid proliferation of applications as new systems come online and investigators piece together optimal strategies for combining the complementary information from these modalities.