TY - JOUR T1 - Similarity and dissimilarity between synaptic density, blood flow and glucose metabolism in Alzheimer’s disease JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 45 LP - 45 VL - 62 IS - supplement 1 AU - Takuya Toyonaga AU - Jayanta Mondal AU - Adam Mecca AU - Ming-Kai Chen AU - Yihuan Lu AU - Mika Naganawa AU - Ryan O'Dell AU - Henry Huang AU - Christopher Van Dyck AU - Richard Carson Y1 - 2021/05/01 UR - http://jnm.snmjournals.org/content/62/supplement_1/45.abstract N2 - 45Objectives: We have proposed PET imaging of synaptic density with the SV2A-targeting radiotracer 11C-UCB-J as a useful tool for studies in Alzheimer’s disease (AD)1,2. Previously, we found that the influx constant K1 showed similar patterns in AD to FDG PET1. Here, we extend that work to investigate the similarity and dissimilarity between brain synaptic density, blood flow (K1) and glucose metabolism in AD. Canonical correlation analysis (CCA) was applied to investigate multivariate imaging relationships and the subject values from multivariate analysis were compared with Mini Mental State Examination (MMSE). Methods: A cohort of 44 participants with mild cognitive impairment (MCI) or dementia due to AD (71.0±7.8 y.o., M:F=22:22) and 17 age-matched healthy subjects (HS) (71.0±7.9 y.o., M:F=8:9) underwent 11C-UCB-J and 18F-FDG PET scans with the high resolution research tomograph (HRRT; 11C-UCB-J: 621.0±162.4 MBq, 18F-FDG: 181.0±5.6 MBq). Individual T1 weighted MR images were analyzed by FreeSurfer (FS) to segment the gray matter (GM) into 83 regions of interest (ROIs). Iterative Yang (IY)3 PVC was performed on dynamic frames using individual FS ROIs. For 11C-UCB-J, simplified reference tissue model 2 (SRTM2) was applied on 0-60 min dynamic data to estimate distribution volume ratio (DVR) and relative influx rate constant (R1) to the cerebellum. For 18F-FDG, 60-90 min standardized uptake value (SUV) images were created and regional SUV ratio (SUVR) to cerebellum was calculated. Principal component analysis (PCA) was applied on each parameter (DVR, R1, SUVR) for dimension reduction and the 1st 4 components were selected for CCA. CCA was applied to image pairs (DVR_R1, DVR_SUVR, R1_SUVR) to investigate linear combinations of ROIs with maximal between-image correlation. ROI weights per image were estimated and these weights were compared to evaluate similarity and dissimilarity. For each comparison, a subject loading value was calculated, and these values were used to assess group differences and to correlate with MMSE. Results: Overall, CCA showed similar ROI weights between image pairs. Correlation coefficient of the ROI weights were 0.90 for DVR_R1, 0.87 for DVR_SUVR, and 0.97 for R1_SUVR (Fig. 1), i.e., the similarities were greatest between flow (R1) and metabolism (SUVR). The ROIs for left and right caudate, left precuneus, and left inferior parietal had larger weights for all combinations, i.e., those ROIs had the most consistent trends between DVR, R1 and SUVR in AD. The variations between DVR and both R1 (UCB-J) and SUVR (FDG) were larger, suggesting that the alteration of synaptic density in AD might be different than that of flow and metabolism. Subject loadings from DVR_SUVR and DVR_R1 analyses provided different performance in detecting group differences than hippocampal DVR. The HS-MCI group difference was more significant with hippocampal DVR, however hippocampal DVR did not distinguish MCI and dementia due to AD. In contrast, CCA loadings showed a significant MCI-dementia group difference (Fig. 2, hippocampal DVR: p=0.32, DVR_SUVR composite value: p<0.001, DVR_R1 composite value: p<0.0005). Also, the subject loadings showed significant correlation with MMSE within participants with AD (Fig. 3, hippocampal DVR: p=0.11, DVR_SUVR: p=0.0024, DVR_R1: p=0.0002). Conclusion: CCA showed overall pattern similarity of DVR, R1, and SUVR, with greatest similarity between FDG SUVR and SV2A R1. Using the CCA-derived subject loadings combining DVR with either SUVR or R1 showed significant MCI-dementia group differences and significant correlation with MMSE suggesting that CCA is useful to extract relevant AD features. Reference: 1. Chen et al., JAMA Neurol., 2018 2. Mecca et al., Alzheimers Dement., 2020 3. Erlandsson et al., Phys. Med. Biol., 2012 ER -