Prolonged latent 'baseline' state of large-scale resting state networks in Alzheimer's disease as revealed by hidden Markov modelling

Abstract Alzheimer's disease (AD) is a progressive neurodegenerative disorder. While resting state fMRI holds great promise in identification of diagnostic markers, how spatio-temporal dynamics of functional networks are reconfigured in AD remains elusive. We employed hidden Markov model to examine the time-resolved information of resting state fMRI data from Alzheimer's Disease Neuroimaging Initiative dataset. Two hundred and ninety-four participants well selected (23 with AD, 54 with mild cognitive impairment and 217 normal controls). We focused on the mean activation map which allows reliable measurement for statistical characteristics of spatial distribution of the latent states. At the time scale of seconds, we detected a 'baseline' state at which all the resting state networks had low activation levels. Moreover, AD patients tended to spend more time on this 'baseline' state and less time on the default mode network states than healthy elderly subjects. The prolonged latent 'baseline' state in AD probably reflects departure of the brain from criticality. Our findings provide important clues that help understand mechanisms underlying the reorganization of large-scale functional networks for AD.