< Back to Article

Imaging of neural oscillations with embedded inferential and group prevalence statistics

Fig 3

iES group analysis: Mapping induced gamma oscillations during visual stimulation.

a) Subspace computation, example subject: (left) values of the quality function f_induced for all the spatial patterns in the MEG data, ranked in decreasing order. The components with the 5 largest values of the quality function were deemed consistent with the tested hypothesis (highlighted with black dots—left, and their sensor topographies shown to the right). This was determined via permutation tests, which yielded , a threshold indicating the minimum value of the quality function for significance (p < 0.05). Note that the number of significant components may vary per subject, as illustrated hereafter. b) Effect prevalence, group level: (left) number of significant spatial components for each subject (K_obs = 17 is the number of participants in this example). The subject illustrated in Panel A is shown in blue; (right) prevalence testing results (as detailed in Materials and Methods) showing the likelihood of the data under a population prevalence γ. γ = .83 is the highest value that can be rejected at p < 0.05 (horizontal dashed line). c) Spatially-filtered signals, example subject: (left) three example trials: the increase in gamma oscillations after stimulus presentation can be readily appreciated visually in the spatially-filtered signals; (right) average time-frequency map across 220 trials: here too, the strong induction of gamma activity is clearly visible. d) Spatially-filtered signals, group level: (left) average wavelet power of spatially-filtered signals in the two time periods of interest (baseline and visual stimulus). Values are expressed in decibels with respect to empty-room MEG recordings, shaded regions are standard errors over subjects; (right) power spectra of the stimulus period in decibels with respect to the baseline period. Thin lines represent single-subject data. e) Subspace correlation maps, example subject: (top) Map of subcorr values in the 3-D source grid, indicating the location of brain regions generating stimulus-induced gamma activity, (bottom) Fisher-z transformed map. f) Subspace correlation maps, group level: (left) a permutation procedure to determine a statistical threshold to apply on the average subcorr scores. The figure shows the histograms of the permuted and observed subcorr values; (right) group-level average subcorr map, thresholded at p < 0.05. The effect confirms the single-subject data shown, and localizes to the occipital visual cortex.

doi: https://doi.org/10.1371/journal.pcbi.1005990.g003