Functional Connectivity in Aperiodic Brain Activity at Rest

1649 

Abstract Submission 

Luc Wilson¹, Jason da Silva Castanheira¹, Sylvain Baillet¹

¹Montreal Neurological Institute, Montreal, Quebec

Luc Wilson  
Montreal Neurological Institute
Montreal, Quebec

Jason da Silva Castanheira  
Montreal Neurological Institute
Montreal, Quebec

Sylvain Baillet  
Montreal Neurological Institute
Montreal, Quebec

Macroscopic brain signals comprise aperiodic and periodic components (Donoghue et al., 2020). Neurophysiological brain activity is thought to covary across the brain in a region- and frequency-specific manner (Brookes et al., 2011). This connectivity has been shown to predict age and fluid intelligence test performance (Jauny et al., 2022). However, conventional measures of functional connectivity cannot resolve the unique contributions of periodic and aperiodic components. Here, we separate dynamic periodic and aperiodic brain signals from resting-state magnetoencephalography (MEG) to identify their unique functional networks. We find aperiodic features of brain signals occupy their own unique functional connectivity profiles (separate from alpha-band activity), and that the aggregate strength of several within- and across-network connections of resting-state aperiodic features are significant predictors of age and fluid intelligence test performance.

We used 2.5 minutes of clean resting-state MEG recordings from 603 participants (298 females; age range = 18 – 88) from the CamCAN dataset (Taylor et al., 2017). Data were preprocessed using Brainstorm (Tadel et al., 2011) before generating cortical source estimates at 15,000 locations through linearly constrained minimum-variance beamforming. From these estimates, we extracted the mean time-series for 200 predefined cortical regions of interest (ROIs) according to the Schaefer-200 atlas, 17-network variant (Schaefer et al., 2018).
We identified the time-varying aperiodic and periodic spectral components of the neural time series at 1-s intervals (2-s windows, 50% overlap) using Spectral Parameterization Resolved in Time (SPRiNT; Wilson et al., 2022). At each time point, we measured the mean alpha-band (8-13 Hz) power and aperiodic-corrected alpha power by subtracting the aperiodic component from the spectrogram. We quantified aperiodic dynamics using spectral parameters (exponent, offset).
To measure functional connectivity across ROIs, we computed Pearson's correlation coefficients between time-varying periodic and aperiodic regional features. We estimated periodic connectivity using mean alpha-band power (before and after removing aperiodic activity), while aperiodic connectivity used spectral exponent and offset parameters resolved in time. We combined within- and across- network edge strengths according to their 17-network designations.
For each edge of the aperiodic networks, we tested whether participant age and Cattell test score were related to edge strength using linear regression models. We corrected for multiple comparisons using a Bonferroni procedure (α = 0.05; N=153).

We observed that alpha-band connectivity (mean r = 0.38) decreased after correcting for aperiodic activity (mean r = 0.24), while aperiodic offset and exponent connectivity remained low (mean r = 0.17, 0.13, respectively) but distinct from the alpha band (Figure 1). Several combined within- and across-network edges in aperiodic connectivity significantly predicted participant age (highest R² = 0.09) and Cattell test score (highest R² = 0.04; Figure 2). Of note, few edges were associated with Dorsal Attention network A or Visual Central network, while a large number of edges were associated with Default Mode network B.

The present study explores the respective contributions of periodic and aperiodic brain signal features to functional connectivity. Aperiodic exponent and offset are thought to reflect excitation-inhibition balance (Gao et al., 2017) and aggregate population spiking (Voytek et al., 2015), respectively. Therefore, functional networks could reflect regional co-fluctuations of these physiological phenomena. Future studies should explore the links between the respective connectivity patterns of a/periodic activity to the topography of meaningful cortical features such cortical thinning with age and neurotransmitter receptor densities.

Higher Cognitive Functions Other

Connectivity (eg. functional, effective, structural) ²

EEG/MEG Modeling and Analysis ¹

Task-Independent and Resting-State Analysis

Aging

Cognition

MEG

Other - Aperiodic; Spectral Parameterization; Functional Connectivity