Voxel-wise estimation of nondisplaceable binding (V_ND) for PET SV2A synaptic density imaging with ¹¹C-UCB-J

Objectives: For noninvasive quantification of PET images by reference region methods, a region in which the total radioligand uptake is representative of the nondisplaceable (i.e. free + non-specifically bound) volume of distribution (V_ND) of the radiotracer is used to determine the specific binding of the PET radiotracer to its target, such as the binding potential with respect to the nondisplaceable compartment BP_ND. V_ND can be determined in vivo using pairs of baseline and blocking scans using specific pharmaceuticals that compete with the PET tracer at its target. Typically, the occupancy (r) and V_ND can be estimated from a Lassen Occupancy Plot, in which the difference of volume of distribution V_T between baseline and blocking conditions is plotted against the baseline V_T. The slope of the best fit line of this data is the estimated occupancy, and the x-intercept represents V_ND. This analysis requires the assumption that both the occupancy and the nondisplaceable binding are uniform across all brain regions. The objective of this work is to examine the assumption of uniform V_ND by using a population of baseline and blocking scans to estimate voxel-wise estimates of V_ND and the specific volume of distribution (V_S) using ¹¹C-UCB-J, a PET radioligand that targets the synaptic vesicle glycoprotein-2A (SV2A).

Methods: Nine healthy subjects underwent a PET scan with ¹¹C-UCB-J before and after a single dose of the SV2A-specific drug Brivaracetam (BRV). Images were acquired on the High-Resolution Research Tomograph (HRRT) and were reconstructed with OSEM (2 iterations, 30 subsets). Time-activity data and arterial input functions were analyzed using a 1-tissue compartment model. Parametric V_T images were created for both baseline and blocking scans and were registered to the Montreal Neurological Institute (MNI) template space. Blood samples were taken at the beginning, middle, and end of the blocking scan to determine drug plasma levels. At each voxel i, using data from baseline and blocking V_T images for each subject j, ordinary least squares methods were used to fit the model in Eq. 1 V_T,i,j = V_ND,i + (1-r_j) [asterisk] V_S,i (Eq. 1) where r_j=c_j/(c_j+IC₅₀) at blocking, where c_j is the plasma BRV concentration during each scan (0 for baseline scans) and IC₅₀ is the known BRV concentration at which 50% of SV2A occupancy is reached. Parametric images were created for V_ND and V_S. V_ND estimates in brain regions were compared to the average extrapolated V_ND found from the corresponding Lassen Occupancy plots.

Results: Across the nine subjects, the average (± SD) Lassen V_ND was 2.65 (± 0.57) mL/cm³. Using the voxel-wise estimation of V_ND and V_S, the estimated V_ND was nonuniform within the brain, ranging from an average of 2.81 (± 0.24) ml/cm³ in the gray matter to 3.38 mL/cm³ in the white matter centrum semiovale region (Table 1). On average, the voxel-wise V_ND estimates in the gray matter regions were consistent with the Lassen V_ND estimates. V_S was also nonuniform, ranging from 1.46 mL/cm³ in the centrum semiovale to an average of 18.2 (± 1.7) mL/cm³ in the gray matter.

Conclusions: The current work proposes a voxel-wise method for estimating the V_ND and V_S of SV2A PET radiotracer, ¹¹C-UCB-J using pairs of baseline and blocking scans. With this method, we report variable V_ND estimates across the brain, which ranged from -6 to +24% different from the average V_ND from the Lassen Occupancy plot. A non-zero V_S estimate in the centrum semiovale, a proposed reference region for ¹¹C-UCB-J, was also observed. Though non-zero V_S in the white matter is not ideal, it is a small fraction of what is observed in the gray matter regions (~8%), and may still be useful for reference region quantification methods. The proposed voxel-wise estimation method may also be useful for other PET radiotracers, to validate or explore new reference regions.