Hidden Markov Model Reveals The Abnormal Brain Dynamics In Patients With Alzheimer’s Disease

Alzheimer’s disease（AD）is the most common progressive neurodegenerative disease in the elderly.The disease will cause patients’ memory,cognitive behavior and social disorders,hinder their daily life and finally endanger their lives.So far,the understanding of AD is still relatively limited,which is difficult to diagnose and treat the disease.In recent years,many studies have tried to use dynamic network analysis to explore the abnormal brain mechanism behind AD.However,the sliding window method and co-activation patterns mode used in most studies have methodological defects.Hidden Markov model（HMM）can decode brain activity in a single frame combined with time sequence information,and is adept in detecting the ’baseline’ state that may be closely related to AD disease,which is helpful for people to have a new comprehension of the abnormal brain dynamics of AD.In this study,we evaluated the repeatability of HMM method and explore its influencing factors.The results showed that HMM can still maintain ideal reliability when the sample size,sampling rate and characteristic dimension change.Then,we utilized the resting state f MRI data of nearly 300 subjects from Alzheimer’s Disease Neuroimaging Initiative dataset and modeled the dynamic changes of spontaneous brain activity with hidden Markov modeling.We found that compared with healthy subjects and mild cognitive impairment patients,AD patients tended to spend less time on two default mode networks related antagonistic states and more time on a ’baseline’ state with moderate-level activation of all networks.In addition,fractional occupancy of the ’baseline’ state was found to be correlated with severity of diseases as indicated by clinical scores.Our results implied that HMM is a method with strong reliability,and its time parameters have ideal test-retest reliability.We found the abnormal ’baseline’ period of resting state networks,probably reflecting the departure from the critical state,may serve as one of the potential neuroimaging biomarkers for AD.