Functional Reorganization in Alzheimer’s Disease: A Network Perspective

Alzheimer’s disease (AD) is a major contributor to the global dementia burden, with prevalence expected to nearly double in Europe by 2050. As clinical symptoms emerge late in the disease course, identifying early functional brain markers of cognitive decline has become a public health priority. Resting-state functional connectivity (rsFC) within the Default Mode Network (DMN) has been consistently linked to episodic memory impairment in AD, but detecting subtle, early disruptions remains challenging. This study investigates whether fine-grained alterations in DMN organization, captured through functional gradient mapping and graph-theoretical analysis, can serve as early indicators of memory decline across the AD continuum. We analyzed multimodal neuroimaging, structural, and proteomic data from 279 participants in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) at baseline and two-year follow-up. Participants were stratified by amyloid status: myloid-positive (A+) and amyloid-negative (A−), to distinguish individuals with early AD pathology. Episodic memory performance (encoding, retrieval, recall) was assessed using the EMBIC Digital Cognitive Biomarker, a validated composite sensitive to early cognitive change. From a gradient-based perspective, DMN dispersion, quantifying variability in functional connectivity across DMN regions, was differentially related to memory depending on amyloid status: in A+ individuals, higher dispersion was associated with poorer memory performance, while in A− individuals, greater dispersion correlated with better memory. Notably, baseline DMN dispersion predicted future memory decline and was associated with longitudinal increases in tau burden, underscoring its prognostic utility. Graph-theoretical analyses provided complementary insights. We examined nodal strength, indicating the total connectivity of a region with its neighbors; participation coefficient, capturing how widely a region connects across different functional modules; and betweenness centrality, reflecting a region’s role as a communication hub based on its presence along shortest paths between other nodes. In A+ individuals, memory was negatively associated with nodal strength and positively associated with participation coefficient in regions such as the dorsomedial prefrontal cortex (DMPFC), middle temporal gyrus (MTG), and inferior parietal lobule (IPL). These patterns suggest that reduced local connectivity, coupled with broader inter-network communication, supports memory in early AD. The left MTG emerged as a region of interest, where lower betweenness centrality and higher participation coefficient predicted better encoding performance, reflecting a distinctive topological signature. An exception was found in a small precuneus region, where encoding correlated positively with nodal strength and negatively with participation, highlighting region-specific dynamics. Together, these findings highlight how functional gradients and graph-theoretical metrics capture complementary aspects of early DMN reorganization in AD. Their integration offers a refined understanding of network-level changes preceding overt symptoms and may support the development of sensitive biomarkers for early detection and intervention.