Alterations in neurocognitive networks in dementing disorders as assessed with simultaneous PET/fMRI

Objectives Functional magnetic resonance imaging (fMRI) studies have reported impaired integrity of resting state networks in Alzheimer’s disease (AD) and frontotemporal dementia (FTD). Among so-called neurocognitive networks are default mode (DMN), salience (SN), and central executive networks (CEN). As neuronal activity is closely linked to glucose consumption, PET with [18F]fluorodeoxyglucose (FDG) represents an attractive tool for mapping neural networks in vivo. Thus, the aim of this combined fMRI/FDG-PET study was to characterize alterations in network integrity in patients with AD and FTD.

Methods Twenty seven patients with AD (MMSE 24.1±2.7), 16 patients with FTD (MMSE 25.5±3.0), and 26 healthy subjects underwent a simultaneous resting state fMRI and FDG-PET on a hybrid PET/MR system. After a standard image pre-processing, data were analyzed using spatial independent component analysis. Thus, a number of networks were extracted in a user-independent manner separately from fMRI and PET data. For PET, we quantified so-called loading coefficients, a degree of network expression in each subject. For fMRI, a goodness-of-fit of each network to a standard network template was calculated. Thus, fMRI- and PET-based indices of network integrity were derived for each network and subject.

Results As compared to healthy subjects, integrity of SN was reduced in FTD, while DMN and CENs were affected in both AD and FTD in fMRI data. In PET data, SN integrity was reduced in FTD, DMN in AD, while CEN was affected in both disease groups. In fMRI data, no significant differences between the AD and FTD groups were found. In PET data, SN integrity was reduced in FTD relative to AD (p<0.001), while DMN integrity was reduced in AD relative to FTD (p<0.05). In a step-wise binary logistic regression, PET indices of SN integrity alone differentiated between the AD and FTD groups with an accuracy of 81.4 % (p<0.005). Correlations between PET- and fMRI-based indices of network integrity were low.

Conclusions Our preliminary results in mildly affected patients with AD and FTD indicate limited disease specificity of neurocognitive networks in general and DMN in particular. Only SN as measured with FDG-PET appeared to discriminate between two patients groups. FDG-PET and fMRI seem to track partly different aspects of network integrity.