Journal of Healthcare Informatics Research Искусственный интеллект
A Multilayer Network-based Framework of Multimodal Brain Connectivity for Alzheimer’s Topological Marker Discovery
Аннотация
Alzheimer’s Disease (AD) is increasingly recognized as a disconnection syndrome disrupting both structural and functional brain networks. Traditional single-layer graph analyses often fail to capture the integrative, multimodal nature of these alterations. We propose a multilayer graph-theoretical framework that integrates structural and functional connectivity and characterizes disease-associated network organization through centrality, clustering, global efficiency, modularity, and multilayer motif analysis. A key contribution is the analysis of recurring higher-order subgraph patterns across layers, complementing conventional global and mesoscale descriptors with localized topological information. Experiments on virtual connectomes derived from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database reveal systematic and statistically significant differences between the ADNI-derived and Healthy Ageing (HA) groups across multiple network measures and multilayer configurations. These findings identify an interpretable set of candidate topological markers characterizing differences in network organization between the two connectomic cohorts. The present study is a group-level feature-discovery analysis; assessment of individual diagnostic performance, clinically actionable thresholds, and utility for earlier disease stages such as mild cognitive impairment requires dedicated predictive validation on independent and longitudinal cohorts.