Graph theory provides a powerful tool to describe quantitatively the topological organization of brain connectivity. A better understanding of network disruption in Alzheimer’s disease (AD) and other neurodegenerative diseases may provide non-invasive biomarkers for dementia differential diagnosis and disease monitoring. This study aimed at investigating functional brain network architecture in late-onset (LO) and early-onset (EO) AD, mild cognitive impairment (MCI), and behavioral variant of frontotemporal dementia (bvFTD).The study involved 122 AD patients, 61 MCI patients, and 51 age-matched controls. 35 EOAD patients were also compared with 29 bvFTD patients and 35 age-matched controls. All subjects underwent 3D T1-weighted and resting state (RS) functional MRI. Network nodes were defined parcellating the AAL atlas (Fig.1A) into 262 regions, using a new parcellation approach which combines the need for high number of equal sized nodes with respecting brain anatomy (Fig.1B). The parcellated atlas was coregistred to the RS subject space (Fig.1C) and the BOLD signal was extracted from each node and correlated among node pairs to obtain functional matrices (Fig.1D). Graph theory analysis was used to measure the global topological properties of functional brain networks and to define brain modules. Differences in regional functional networks among groups were investigated using Network-based statistic.While controls showed high-densely connected modules, AD groups and bvFTD patients showed a loss of long-distance intra-module connections, involving all modules in AD and the fronto-parietal-parahippocampal module in bvFTD (Fig.2). Regardless the age of onset, AD patients showed altered global network measures (lower network degree and clustering coefficient, and longer path length) compared to controls and to the other patient groups. Although MCI patients did not show global network alterations, they were characterized by a decreased regional functional connectivity in the fronto-parietal connections compared with controls. A decreased regional functional connectivity was prominent in the parieto-occipital connections in EOAD and in the fronto-temporal-parietal connections in bvFTD patients (Fig.3).Global graph properties of brain networks are severely altered in AD, while they are relatively maintained in the other patient groups, thus suggesting that they are promising in distinguishing EOAD from bvFTD patients. Furthermore, the fronto-parietal connectivity disruptions in MCI patients could reflect an early marker of the disease. Graph analysis is promising to detect functional changes in the early phase of neurodegenerative diseases and to serve for a prompt differential diagnosis among dementia syndromes.