Abstract
138
Objectives: Head motion during PET sessions may confound identification of regions of neuroreceptor activation with PET. SPM results may also be affected by the accuracy of binding potential (BP) maps which may differ with modeling approaches. Thus, we examined whether combinations of the two factors improve the ability of SPM to identify activation clusters on a [11C]raclopride neuroreceptor activation study. Methods: Sixteen young normal subjects (age: 30.2± 9.1) underwent two 90-min [11C]raclopride PET sessions while viewing videos of natural (Scan 1) and drug-use (Scan 2) scenes. PET images were prepared without (noHMC), and with frame-by-frame head motion correction (HMC) using the mutual information theory (MIT-HMC) and a new partial-volume correction-based method (PVC-HMC). BP maps were constructed by three published methods listed in the Table. Each BP map was spatially normalized to a standard brain and smoothed with a 10 mm Gaussian kernel to compare BP maps between Scan 1 and 2 using SPM2. Results: All combinations showed one cluster at [24, 16, z] where z ranged from -2 to 6 (Table). All modeling approaches had the most significant clusters with PVC-HMC. RTGA showed weaker and smaller clusters than MRTM2 and BPIT. The two approaches with PVC-HMC showed clusters at a p<0.05 with false discovery rate (FDR) correction. Conclusions: This study confirmed that the outcome of voxel-wise statistical method (SPM) depends on the selection of the head motion correction and BP map generation methods. The findings warrant further improvements in the two factors beyond the combinations of the PVC-based HMC with MRTM2 or BPIT modeling approaches that yielded the best results in this study.
Research Support (if any): Grant support: AA12839; MH078175; K24 DA00412
- Society of Nuclear Medicine, Inc.