Optimizing the sensitivity of lp-ntPET to detect transient DA release induced by smoking in PET data.

Objectives The lp-ntPET model applied to dynamic raclopride PET has proved effective in detecting transient dopamine (DA) release at the voxel level in humans[1, 2]. HYPR is a PET denoising algorithm that smoothes in space without sacrificing temporal resolution[3]. Our goal is to optimize study design and image processing procedures to maximize sensitivity to transient smoking-induced DA release with a fixed, low false positive rate. We will compare the performance of bolus and bolus plus infusion (B/I) paradigms with and without HYPR for a constant delivered radioactivity dose.

Methods 1. "Null" simulation. To determine a cluster size threshold to maintain low false positives, 100 "null" datasets each for bolus and B/I designs were simulated. For each dataset, a ‘Significance Mask’ was created following Kim et al[1] of the voxels that appeared to contain a significant DA signal. The cluster-size threshold was selected so 90% of the significance masks were completely free of random clusters. 2. "Smoking" simulation. To determine the sensitivity, 100 “smoking” datasets were simulated with clusters of transient DA release starting at 35min into the scan. The cluster-size threshold determined in step 1 was applied to the ‘Significance Masks’ to generate the ‘Final Significance Masks’, which were summed to create sensitivity maps.

Results The sensitivity of the bolus design was nearly as good as that of B/I for the same delivered dose. HYPR increased sensitivity in both bolus and B/I, although the cluster size threshold with HYPR was higher than without.

Conclusions A bolus design requires less synthesized activity vs. B/I to achieve the same delivered dose. Thus, bolus studies are less likely to fail for lack of activity. In case of limited tracer activity, the bolus design with HYPR processing yields superior sensitivity to detect transient DA activation in PET data.

Research Support R21DA032791, P50DA033945