Spectral graph modeling reveals global slowing of neurophysiological network transmission in Alzheimer’s disease

Abstract

AbstractBackgroundAlzheimer’s disease (AD) is the most common form of dementia, progressively impairing memory and cognition. While various neuroimaging studies have revealed functional network abnormalities in patients with AD, how these relate to aberrant neuronal circuit mechanisms remains unclear. In order to answer these questions, we employed a spectral graph‐theory based model (SGM) to identify abnormal biophysical markers of neuronal activity in AD.MethodSGM is an analytic model that describes how the long‐range fiber projections in the brain mediate and couple the excitatory and inhibitory activity of local neuronal subpopulations. Unlike other coupled models, the SGM is linear, available in closed‐form, and parameterized by a small set of global parameters. This facilitates their rapid and unambiguous inference which we performed here on a well characterized clinical population of patients with AD (N = 88) and a cohort of age‐matched healthy controls (N = 88). We estimated SGM parameters that best captured the regional magnetoencephalography (MEG) frequency spectra of patients with AD and age‐matched controls.ResultSGM was able to capture the regional MEG frequency spectra of patients with AD and healthy controls, specifically, the abnormal slowing of neural oscillations in AD (Fig 1A). Patients with AD have significantly elevated long‐range excitatory neuronal time constant (tG) compared to controls (p < 0.001). Time constant tG is also the most important feature for accurate classification of patients with AD from controls (Area Under Curve = 0.85; Fig 1B). Lastly, higher tG is negatively correlated with global cognitive decline score (p = 0.02; Fig 1C).ConclusionThese results indicate that a global impairment in the long‐range excitatory neuronal activity is associated with abnormal spectral signatures and cognitive deficits in AD. Intriguingly, our work is able to recapitulate the spatial and spectral patterns of AD‐related functional activity without introducing any spatial heterogeneity whatsoever; indeed, the SGM model is entirely global and spatially‐invariant. This raises the possibility that a global increase in the long‐range excitatory time constant might be a sufficient factor underlying observed spatiotemporal alterations of neuronal activity in AD. Our findings provide new insights into potential mechanistic links between abnormal neural oscillations and their cellular correlates in AD.