Premium
fMRI complexity is associated with tau‐PET and cognitive decline in Alzheimer’s disease
Author(s) -
Jann Kay,
Kim Hosung,
Albrecht Daniel,
Wang Danny J.J.
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.045411
Subject(s) - parahippocampal gyrus , statistical parametric mapping , neuroimaging , psychology , neuroscience , alzheimer's disease , cognitive decline , magnetic resonance imaging , medicine , dementia , temporal lobe , pathology , disease , epilepsy , radiology
Background Amyloid‐PET is considered an early marker for the preclinical stage of AD, while the neurofibrillary tangle pathology detected with tau‐PET imaging correlates more closely with neuronal injury and cognitive decline. However, PET scans are expensive and involve radiation exposure. Recently, entropy measures have been explored as indices of the complexity of rs‐fMRI time‐series. Reduced entropy values were associated with aging, APOE ɛ4 genotype and cognitive decline in AD. Here we use complexity of BOLD signals as an index of the information processing capacity of regional neuron populations, and test the hypothesis that it is sensitive to tau‐related neuronal injury and cognitive decline in the AD processes. Method Data used in this study were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Our cohort consisted of a sample of 50 subjects form ADNI‐3 (age = 72.4±8.2, 19M/31F, 25CN, 19MCI, 6AD) with a T1 structural, a tau‐PET (tracer: 18F‐AV1451) and two fMRI scans. FMRI data underwent standard preprocessing before computing Multi‐Scale Entropy (MSE). Parametric SUVR images were calculated with a cerebellar reference region. Correlations between SUVR tau‐PET and MSE were calculated for 195 ROIs (Craddock atlas parcellation) including age, gender and regional gray matter volume as covariates. To evaluate if MSE and PET can be used to classify AD from CN we trained and tested a random‐forest classifier ( via 10‐fold cross‐validation). Result We found significant negative correlations between low frequency MSE and tau‐PET measures in hippocampus, parahippocampal gyrus and posterior cingulate cortex (Figure 2). The classification model revealed that MSE and PET have similar prediction accuracy and sensitivity (Table Figure 3A). The most informative areas for the prediction revealed a large degree of overlap between MSE and tau‐PET (Figure 3B). Conclusion Correlation and classification results for MSE in relation with tau‐PET revealed areas in PCC, temporal and parietal lobes. These areas are implicated in tau pathology and cognitive decline. Hence, these overlapping findings support our hypothesis that the complexity of rs‐fMRI is associated with regional tau protein accumulation measured by PET and could provide a marker for information processing capacity of regional neuron populations and prediction of cognitive decline in the AD processes.