Lifespan Mapping of EEG Source Spectral Dynamics with Xi-AlphaNET
We formulate a new class of parametric, multivariate, and structurally informed spectral components model of the EEG, the Xi-AlphaNET, that allows us to map the Lifespan of EEG source spectral dynamics across a large data set of EEG cross spectrum at high spatial resolution. This approach accurately estimates source spectral components and effective connectivity through the use of biophysical modeling while maintaining computational efficiency, as confirmed by simulation benchmarks. We are able to analyze source dynamics with a resolution of 8,003 voxels from the HarMNqEEG dataset, which includes scalp EEG cross-spectrum tensors collected from 1965 subjects across 9 countries, using various devices and accounting for different age groups. Our findings indicate that the Bayesian Model Inversion of the Xi-AlphaNET allows to map Lifespan of conduction delays that follows a U-shaped trajectory, which contrasts with independently recorded myelin concentration measurements. Moreover, we assess the spatiotemporal distribution of spectral components, revealing that the aperiodic or fractal component has an isotropic spatial distribution on the cortical surface. While the generator's spectral peak in the alpha band, i.e., alpha-rythms, is localized on the visual areas of the brain. Using a Zero Inflated Gaussian model, our findings indicate that the mode frequency that characterizes the alpha-rythms or Peak Alpha Frequency shows an inverted U-shaped trajectory for both hemispheres across the Lifespan and a spatial gradient of zero inflation in PAF across the cortex that flattens the trajectory from posterior to frontal areas. We provide both the code of the Xi-AlphaNET and the source solution of the spectral dynamics for the HarMNqEEG.