Premium
ICA‐derived sources of synaptic density PET ([ 11 C]UCB‐J) relate to cognitive impairment severity in Alzheimer’s disease
Author(s) -
Fang Xiaotian T.,
Mecca Adam P.,
Naganawa Mika,
O'Dell Ryan S.,
Chen MingKai,
van Dyck Christopher H.,
Carson Richard E.
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.041197
Subject(s) - clinical dementia rating , dementia , alzheimer's disease , hippocampus , neuroscience , entorhinal cortex , psychology , independent component analysis , cognition , population , cerebellum , audiology , cognitive impairment , pathology , medicine , disease , artificial intelligence , computer science , environmental health
Background Synaptic loss is a primary pathology in Alzheimer's disease (AD) and correlates best with cognitive impairment as found in postmortem studies. Previously, we observed in vivo synaptic density loss with [ 11 C]UCB‐J PET (radiotracer for synaptic vesicle protein 2A, SV2A) in hippocampus and throughout the cortex. Here, we used independent component analysis (ICA), a data‐driven method that identifies independent sources (networks) by linearly unmixing the observed signal into maximally independent components. The aim is to apply ICA, which does not use subject group information, to synaptic density PET data to identify brain networks associated with cognitive deficits in AD. Method [ 11 C]PIB PET was performed to determine brain amyloid status. A structural T1‐weighted MR was acquired to exclude structural abnormalities and for PET co‐registration and analysis. [ 11 C]UCB‐J binding to SV2A was measured in 38 AD (24 dementia, 14 mild cognitive impairment) and 19 cognitively normal (CN) participants. [ 11 C]UCB‐J distribution volume ratio values were calculated using SRTM2 (population‐fixed k 2 ') with whole cerebellum as reference, registered into MNI space and smoothed (8‐mm Gaussian). Based on previous SV2A ICA, 18 independent components (IC) were extracted. Subject loadings per IC were compared between groups using unpaired t‐tests. Pearson's correlations were used to assess relationships between loading weights and cognitive measures: logical memory II (LMII), Rey auditory verbal learning test (RAVLT‐delay), clinical dementia rating sum of boxes (CDR‐sob), mini‐mental state examination (MMSE). Result For 8 ICs, we observed significant differences in loading weights between CN and AD groups, including a component (IC01) located in hippocampus and temporal lobe (Figure 1). Within the AD/MCI group, loading weights for IC02, located in the right parietal‐temporal lobe (Figure 2), correlated with LMII (R 2 : 0.13, p =0.025), and RAVTL‐delay (R 2 : 0.13, p =0.027). For IC03, located in the left parietal hemisphere (Figure 3), loading weights correlated with CDR‐sob (R 2 : 0.23, p=0.0028) and MMSE (R 2 : 0.13, p =0.024). Conclusion We demonstrated that ICA could define coherent spatial patterns of synaptic density. Furthermore, commonly used cognitive measures correlate significantly with loading weights for two of such networks within only the AD/MCI group. Further studies will explore whether ICA‐based networks might be more sensitive than conventional ROI‐based analysis.