Abstract
1627
Objectives: Resting-state networks (RSN) are systems of brain regions anatomically separated but functionally connected during rest [1]. First identified in functional MRI studies based on blood-oxygen-level-dependent (BOLD) signal, RSNs have also been found with 18F-FDG [2,3]. It has been proposed that RSNs represent intrinsic brain connectivity networks that reflect global functional organization during active as well as rest states. 11C-UCB-J is a PET tracer that binds to and images synaptic vesicle protein 2A (SV2A). It is proposed as a measure of synaptic density and thus, it is a promising method for investigating RSNs through examination of regional covariance in synaptic density that might be expected to result from coherent patterns of neural activity. The objective of this work is to identify RSNs in 11C-UCB-J PET data.
Methods: 11C-UCB-J was synthesized at the Yale PET Center as described previously [4]. Healthy subjects (n=72, 27 female/45 male, 45±17y) were administered an i.v. bolus injection of 11C-UCB-J over one min (536.77±180.60 MBq, specific activity: 109.24±36.75 MBq/nmol) and received one dynamic PET scan (HRRT, Siemens) with a reconstructed image resolution of approximately 3 mm. MR scans were also collected (3T Trio, Siemens) for PET coregistration and subsequent registration into Montreal Neurological Institute (MNI) space. Parametric volume of distribution (VT) images were generated with a 1T compartment model using the metabolite-corrected arterial plasma curve. Based on the VT image, BPnd maps were calculated with the centrum semiovale as reference region (i.e. (VTvoxel/VTref)-1), then smoothed with a Gaussian kernel of 8 mm FWHM. Independent component analysis (ICA) was used to identify RSNs [5]. ICA extracts maximally independent components (IC) from an unknown linear mixing matrix of random non-Gaussian vectors containing independent source signals that generate the observed signals (x). ICA solves for y = Wx, i.e., the un-mixing matrix (W) of source maps (y) and is often used to identify RSNs in fMRI data. For exploratory ICA of 11C-UCB-J, parametric images were loaded into the source-based morphometry (SBM) toolbox of the Group ICA of fMRI Toolbox (GIFTv4.0b). BPnd images were thresholded with a relative mean threshold of 0.7. In order to identify RSNs, exploratory ICA was performed with an IC number of 15, similar to previous fMRI analyses. Component spatial loading values were z-transformed and visualized using a z score threshold of 2 to determine the most prominent structures. The resulting maps were visually inspected and compared to previous resting-state ICA findings for functional classification.
Results: Of the 15 ICs, 8 appeared spatially consistent with established RSNs. These can be characterized as primary visual posterior, default-mode network, secondary visual, cerebellar, executive control, basal ganglia, mesial parietal/prefrontal, and primary visual anterior. Preliminary analysis found significant age and gender effect with a few ICs, however these results require cross-validation to avoid Type I error.
Conclusions: With an ICA-based approach, we demonstrate the extraction and identification of RSNs from 11C-UCB-J PET data in a healthy population. These data suggest activity in functional brain networks may be related to coherent, network-based changes in synaptic density, though further multimodal validation with fMRI data is warranted. References: 1. Raichle et al. 2001, PNAS, 98:676-82. 2. Di et al. 2012, Brain Connect, 2:275-283. 3. Savio et al. 2017, J Nucl Med, 58:1314-1317. 4. Nabulsi et al. 2016, J Nucl Med, 57:777-784. 5. Smith et al. 2009, PNAS, 106:13040-13045.