BackgroundFrontotemporal dementia (FTD) is a complex neurodegenerative disorder encompassing heterogeneous subtypes, including behavioral variant frontotemporal dementia (BV-FTD), semantic variant frontotemporal dementia (SV-FTD), and progressive non-fluent aphasia frontotemporal dementia (PNFA-FTD). Unraveling the shared and distinctive brain module organizations among these subtypes is critical for unraveling the underlying neural basis of the disease. This study aims to explore brain module organization in FTD subtypes, seeking potential biomarkers and insights into their pathophysiology.MethodsResting-state functional magnetic resonance imaging data were obtained from the Frontotemporal Lobar Degeneration Neuroimaging Initiative, comprising 41 BV-FTD, 32 SV-FTD, 28 PNFA-FTD, and 94 healthy controls, following exclusion of participants with excessive head motion. Individual functional brain networks were constructed at the voxel level of gray matter and binarized with a 1% density threshold. Using predefined brain modules, we computed the modular segregation index (MSI) for each module, analyzed intermodular and intramodular connections to identify driving modular connections, and calculated the participation coefficient (PC) to detect regions with altered nodal properties associated with module integrity. A machine learning algorithm was employed for FTD subtype classification based on these matrices. Correlations between modular measures and clinical scores in each FTD subtype were examined.ResultsDistinct brain module organizations were observed across FTD subtypes, with lower MSI in the subcortical module (SUB), default mode network (DMN), and ventral attention network (VAN) in both BV-FTD and SV-FTD. Specifically, only BV-FTD exhibited disruption in the frontoparietal network (FPN). Notably, the bilateral fusional gyrus, left orbitofrontal cortex, left precuneus, and right insular thalamus showed significant group effects on PC, indicating altered nodal properties associated with module integrity. Our machine learning achieved a multiple classification accuracy of 85%. Correlations between specific network alterations and clinical variables in each FTD subtype were also identified.ConclusionsThese findings illuminate the diverse brain module organization in different FTD subtypes, offering insights into potential neurobiological differences that underlie the clinical heterogeneity of the disease. Regions with altered modular properties may serve as valuable biomarkers for early diagnosis and disease monitoring. Furthermore, understanding disruptions in modular connectivity provides valuable insights into the neuropathological mechanisms of FTD subtypes, paving the way for targeted therapeutic interventions.
