Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by the accumulation of toxic misfolded proteins, which are believed to have propagated from disease-specific epicenters through their corresponding large-scale structural networks in the brain. Although previous cross-sectional studies have identified potential AD-associated epicenters and corresponding brain networks, it is unclear whether these networks are associated with disease progression. Hence, this study aims to identify the most vulnerable epicenters and corresponding large-scale structural networks involved in the early stages of AD and to evaluate its associations with multiple cognitive domains using longitudinal study design. Annual neuropsychological and MRI assessments were obtained from 23 patients with AD, 37 patients with amnestic mild cognitive impairment (MCI), and 33 healthy controls (HC) for 3 years. Candidate epicenters were identified as regions with faster decline rate in the gray matter volume (GMV) in patients with MCI who progressed to AD as compared to those regions in patients without progression. These epicenters were then further used as pre-defined regions of interest to map the synchronized degeneration network (SDN) in HCs. Spatial similarity, network preference and clinical association analyses were used to evaluate the specific roles of the identified SDNs. Our results demonstrated that the hippocampus and posterior cingulate cortex (PCC) were the most vulnerable AD-associated epicenters. The corresponding PCC-SDN showed significant spatial association with the patterns of GMV atrophy rate in each patient group and the overlap of these patterns was more evident in the advanced stages of the disease. Furthermore, individuals with a higher GMV atrophy rate of the PCC-SDN also showed faster decline in multiple cognitive domains. In conclusion, our findings suggest the PCC and hippocampus are two vulnerable regions involved early in AD pathophysiology. However, the PCC-SDN, but not hippocampus-SDN, was more closely associated with AD progression. These results may provide insight into the pathophysiology of AD from large-scale network perspective.
Bayesian Multivariate Growth Mixture Modeling of Longitudinal Data: An Application to Alzheimer’s Disease Study
