Extended Data Fig. 4: Individual participant distributions of hippocampal FC alongside cycles of seizure risk.
From: Hippocampal network activity forecasts epileptic seizures
(a) Model performance (R2) of phase-encoding generalized linear model (PE-GLM) relating multidien IEA phase as a predictor of local (left, red) and long-range (right, blue) functional connectivity across participants (n = 15). True model performance is displayed in comparison to a permutation-based null distribution; bold, mean; shaded, SEM; dashed 99.9% confidence interval. IEA phase predicts functional connectivity changes across time significantly better than the permutation-based surrogate model across a broadband range of spectral frequencies for both local and long-range connections (one-sided permutation test; p < 0.05, Bonferroni corrected) across participants. (b-f) Distribution of local (red) and long-range (blue) functional connectivity per study participant (y-axis) over two cycles of the multidien IEA cycle (x-axis) for (b) delta band, (c) theta band, (d) alpha band, (e) beta band, and (f) gamma band. Color intensity shows peak FC strength within the frequency band in an individual participant (purple colors indicate overlapping local and long-range connectivity). Contour lines reflect 99% confidence intervals generated by a permutation-based surrogate data generation model that preserves amplitude distributions while shuffling the temporal order of phase within the signal. Participant order sorted by hippocampal FC preference to the trough of the cycle. Consistency in the relationship between FC and IEA cycle phase is demonstrated by uniform color intensity across rows (participants) at discrete columns (cycle phases). Thick bordered boxes are colored based on functional connectivity type found to be significant at the population-level (Fig. 1e).
