Fig. 1

Construction of power ratio and functional networks from multivariate scalp EEG recordings. ai Example EEG data from re-referenced 18 channels (broadband, 0.5 - 50 Hz) according to the bipolar “double banana” montage. Filtered and unfiltered data are divided into 2 s epochs. aii From unfiltered data power spectra are calculated for each channel using the multitaper method. aiii The ratio of power spectra are obtained from the power spectra of the posterior four derivations (T5-O1, P3-O1, P4-O2, T6-O2) divided by the anterior four derivations (Fp1-F7, Fp1-F3, Fp2-F4, Fp2-F8). Shown here is the mean posterior/frontal power spectra ratio to illustrate the properties of the peak alpha-ratio statistic. b For each channel pair filtered data (0.5 - 50 Hz) from 2 s epochs are used to calculate the cross-correlation. Two example traces for Fp2-F8 and T4-T6 show a correlation here with maximal coupling at a time lag of −50 ms. The significance of the maximum absolute value of the cross-correlation (blue circle) is determined using an analytic procedure (see Methods). c Example binary coupling networks derived from four 2-s epochs. Significant electrode coupling is represented with an edge. These networks are averaged, resulting in a weighted coupling network for each subject. These are then compared against bootstrapped edge weight distributions in (d). d To create bootstrapped edge weight distributions, surrogate networks mirroring the original datasets are created by randomly sampling functional networks with replacement from all epochs of all subjects of both groups. Original ASD and control edge weights are compared to the surrogate edge weight distributions, and edges most significantly outside the distribution (p < 1/100,000) are selected to make a mask of highly significant edges. This mask is used to select the edges with the greatest difference between the ASD and control groups