Optimization of the white matter reference region for synaptic density imaging with ¹¹C-UCB-J PET

Objectives: The new PET tracer ¹¹C-UCB-J, targeting synaptic vesicle glycoprotein 2A (SV2A), has the potential to become a general biomarker for synaptic density evaluation [1]. One tissue compartment modeling applied to dynamic 4D PET data provides high quality images of the volume of distribution (V_T) and tracer delivery (K₁). To estimate binding potential of the tracer, a reference region with only non-displaceable activity is needed. Based on previous investigations, the centrum semiovale (CS) emerged as a prime candidate for the reference region [2] A good reference region would result in low V_T value, small inter-individual variation, and minimal displaceable activity, which means that the V_T values in CS should be consistent between a baseline scan and a blocking study. However, our previous study, where a conventional CS region-of-interest (ROI) created from the AAL template was used, presented higher VT values of the CS region in baseline than blocking scans. It could be due to the spill-in effect from the gray matter regions where contained high uptake due to specific binding. To minimize such effect, in this study, we optimized the ROI choice for CS.

Methods: New CS ROIs for each subject were created using the individual MRI image. The MRI underwent probability mapping of tissue class using SPM12. To detect the central white matter (WM) regions, far from gray matter to prevent spill-in effect, we smoothed the WM probability map with a 10-mm full-width-half-max Gaussian filter. The smoothed WM probability map was masked by an AAL-template CS ROI in the subject space. Finally, within the masked map, six new CS ROIs of different sizes were created by choosing the voxels with the highest probability values with 1, 2, 3, 4, 5, and 6 mL in volume, respectively. An additional CS ROI was created in AAL template space using the average of the 2-mL CS ROIs from healthy subjects registered to the template space. In total seven new CS ROIs were generated for 17 female and 30 male healthy subjects. Each subject underwent 60-min ¹¹C-UCB-J dynamic scan on the HRRT. In addition, 12 blocking scans were performed using the SV2A targeting agent levetiracetam or brivaracetam. The 7 new CS ROIs and a conventional AAL CS ROI were evaluated on each individual V_T image.

Results: The V_T mean ± inter-subject standard deviation (SD) of the conventional CS ROI was 4.76 ± 0.6 mL/cm³, higher than the V_T values using the new ROIs, which ranged from 4.60 (6-mL) to 4.33 (1-mL) (see table). The inter-subject SD and coefficient of variance (CoV) were similar for all ROIs. The volume of the new CS ROI in AAL space was 2 mL, which became 1.28 ± 0.19 mL after being warped into the individual MR space. Note, the volume of conventional CS ROI was 3.46 ± 0.53 mL. Interestingly, the new CS ROI in AAL space had the lowest mean V_T value. The V_T values of all new CS ROIs were significantly lower than that of the conventional CS ROI (p < 0.01). When the new ROIs were applied to the blocking scans, the V_T reduction from baseline scan was significantly smaller for all new ROIs compared with conventional CS ROI (p < 0.003).

Conclusion: All the newly proposed centrum semiovale ROIs yielded lower V_T in healthy subjects than the previous ROI. Furthermore the lowest mean V_T in CS resulted from using new CS ROI in AAL template space. This new CS ROI should provide improved quantification of the reference region for synaptic density PET imaging with ¹¹C-UCB-J. References: 1. Nabulsi et al., 2016, JNMMI, 57:5:777-84. 2. Finnema et al., 2016, Sci Transl Med, 8:348:348ra96.