Structural Connectivity Analysis in Cognitive Decline: Insights from Graph Theory and Mass-Spring Modeling.

The landscape of cognitive states and their underlying neurobiological mechanisms has been significantly illuminated through advancements in neuroimaging and computational modeling. This study introduces an integrated approach that harnesses network analysis and machine learning techniques to characterize and differentiate cognitive groups—Normal Control (NC), Mild Cognitive Impairment (MCI), and Alzheimer’s Disease (AD). Structural networks are formulated and analyzed based on diffusion tensor data through a fusion of graph theory and mass-spring model methodologies. Notably, features extracted from both graph theoretic and mass-spring model computations drive a two-step framework. This process commences with a random forest-based feature extraction, followed by a support vector-based classification approach, culminating in an impressive accuracy of 82.7% for classifying individuals across cognitive groups, with an AUC of 0.893. This study significance is underscored by the pressing need for enhanced cognitive impairment detection and differentiation strategies. The identified features offer nuanced insights into the intricate interplay among brain structure, dynamics, and cognitive function, thereby bridging gaps in our understanding of cognitive decline and neurodegeneration. By fortifying our diagnostic repertoire and facilitating personalized interventions, this research paves the way for refined clinical practices.