Fig. 3

A, B, C: Statistical significance testing of difference in model parameters between AD and controls, with age as a covariate. Distribution of A: long-range parameters \(\tau _{\textrm{G}}\) (long-range excitatory time constant), v (speed), and \(\alpha\) (coupling constant); B: local time constants \(\tau _{\textrm{e}}\) (excitatory) and \(\tau _{\textrm{i}}\) (inhibitory); and C: local neural gains \(g_{\textrm{ii}}\) (inhibitory gain) and \(g_{\textrm{ei}}\) (gain of signals from the coupling between excitatory and inhibitory neurons). P-values are reported after correcting for multiple testing using a Bonferroni correction. \(*\): \(p<0.05\), \(***\): \(p<0.001\). D, E, F: Univariate associations of D: \(\tau _{\textrm{G}}\), E: \(\tau _{\textrm{e}}\), and F: \(g_{\textrm{ii}}\) with MMSE in patients with AD. G, H: Classification of AD vs controls with a random forest classifier with SGM parameters and age as features of the classifier. G: ROC curve for classification of AD versus controls. H: Feature importance plot of SGM parameters. I: Spectral correlations when optimizing for only certain model parameters while keeping the others fixed at the average of the optimized model parameters of both AD and controls. “None” implies that all the model parameters were fixed at the average. The second point on the x-axis with the label \(\tau _{\textrm{G}}\) implies that only \(\tau _{\textrm{G}}\) was allowed to be optimized while the other model parameters were fixed at the average values. The third point on the x-axis with the label \(\tau _{\textrm{e}}\) implies that both \(\tau _{\textrm{G}}\) and \(\tau _{\textrm{e}}\) were allowed to be optimized while keeping the other model parameters fixed at the average values. All the subsequent points on the x-axis correspond to similarly including more model parameters in optimization, based on their importance in the classification of AD vs controls