Altered brain-ventricle coupling modes over Alzheimer's disease progression detected with fMRI
The temporal coordination between fMRI signals in gray matter and ventricular spaces represents an emerging yet poorly understood dimension of brain physiology with potential relevance for neuroprotection. Here, we analyzed 2177 resting-state fMRI scans from 880 participants spanning the cognitive spectrum from normal aging to dementia, using a unified phase-coupling approach that incorporates fluctuations in both brain tissue and ventricular space. We identify distinct modes of Brain-Ventricle (BraVe) coupling occurring over time, and compare their temporal occupancy across cognitive status groups, neurodegenerative biomarkers and cognitive scores. In the most prevalent BraVe mode, signals across the entire brain evolve in anti-phase with ventricular signals, occurring less frequently with cognitive decline (p < 10-7, g > 0.3). In the remaining BraVe modes, specific cortical regions temporarily align with ventricular fluctuations, shaping canonical resting-state networks at the cortical level. Two BraVe modes overlapping with the Default Mode Network and the Frontoparietal Network, occur less frequently in dementia (p < 10-8, g > 0.35), with the latter correlating significantly with FDG-PET glucose metabolism (r = 0.20, p = 3.65x10-5). Conversely, a mode overlapping with the Limbic Network increases in occupancy with cognitive decline and associates with APOE {varepsilon}4 status (r = 0.11, p = 1.30x10-6). BraVe mode occupancy correlates more strongly and broadly with cognitive function than with disease-specific molecular biomarkers, with between 16 and 27 cognitive and functional variables per mode surviving strict Bonferroni correction, spanning global cognition, episodic memory, executive function, and everyday functioning. These findings reveal that the transient phase shifts underlying resting-state network dynamics are not arbitrary: when cortical subnetworks shift out of phase from the global brain signal, they align with ventricular signals, suggesting that resting-state functional connectivity and brain-ventricle coupling may be two perspectives on the same underlying physical process, one that is directly linked with cognitive function.