Table 1.
Demographic and clinical characteristics.
Table 2.
Characteristics of channel zones Z.
Fig 1.
Neuroanatomical representation of resected brain tissue (RBT, panel a), intracranial electrode localization (b) and relative contribution of each contact to the normalized node strength of the surrogate corrected mutual information matrix M (c) for patient I-2 (first seizure).
The spheres in panel c are centered at the positions of the intracranial electrode contacts. Their volume is proportional to the peri-ictal channel-wise mean of the node strength. The implantation scheme in this patient was fully symmetric. For simplicity, only electrodes in the left hemisphere are shown in panel c. The color code is as follows: red, channels included in the RBT; blue, channels belonging to the SOZ; magenta, overlap OVL, i.e. channels that were resected and belonged to the SOZ; black, channels NON that neither belonged to the RBT nor to the SOZ. Channel labels are: TPL, temporo-polar left; TBL, temporo-basal left; TLL, temporo-lateral left; DEL, depth electrode left. A movie showing the contribution of all four measures on the implantation scheme of the left hemisphere in 3D is available in the supplementary material (S1 Movie).
Fig 2.
Peri-ictal evolution of quantifiers derived from the normalized node strength of the surrogate corrected mutual information matrix M of iEEG signals.
The seizure starts at time point zero. Visually determined seizure onset and termination are indicated by vertical lines in the panels e and j. iEEG channels recording from RBT are indicated in red, the SOZ in blue and the overlap of both in magenta on the very left. Channels belonging to none of these zones are indicated in black. The color-scale figure in panel t shows the temporal evolution of each iEEG channel's normalized node strength of M. The vertical bar plots in panels p, q, r and s display the mean channel contribution in each of the following phases: pre-ictal, early ictal, late ictal and post-ictal. All these bar plots are scaled for optimal display. The bars are light shaded, whereas the standard error of the mean contribution is displayed in full color. The temporal evolution of salient (white) and normal channels (black) is shown in panel o. The vertical bar plots in panels k, l, m and n show the channel-wise mean prevalence of salient values in the four peri-ictal phases. For better comparison, here, all bar plots are shown in the same range 0 to 1.1. Panel e shows the temporal evolution of the fraction of salient values falling into the RBT (red) and the SOZ (blue). Panels a, b, c and d show the means over all four zones Z = {RBT,SOZ,OVL,NON} during the four peri-ictal phases. The 95% confidence interval for the mean is displayed as whiskers. The negative logarithm of the probability for randomly finding the observed or a larger amount of salient channels in the RBT (red), the SOZ (blue) and the overlap OVL of both (magenta) is shown in panel j as a function of time and in panels f, g, h and i as mean over the four peri-ictal phases. The 95% confidence interval for the mean is displayed as whiskers and the horizontal line indicates an approximation to the significance threshold -log10(0.05) = 1.30 for temporal means.
Fig 3.
Same neuroanatomical representation as in Fig 1 but for the first seizure of patient IV-1.
The color code is as follows: red, channels included in the RBT; blue, channels belonging to the SOZ; black, channels NON that neither belonged to the RBT nor to the SOZ. In contrast to patient I-2 in Fig 1, there was no overlap OVL between SOZ and RBT. Channel labels are: FAL, frontal anterior left; FLL, frontal lateral left; TLPL, temporo-lateral to polar left; TBLL, temporo-basal left; TLPBL, temporo-lateral to parieto-basal left; TLCL, temporo-lateral to cranial left. A movie showing the contribution of all four measures on the implantation scheme in 3D is available in the supplementary material (S2 Movie).
Fig 4.
Same as Fig 2 but for the first seizure of patient IV-2.
For time steps where no salient channels were identified, the quantifiers and
are undefined, leading to discontinuities of their respective time courses (panels e and j).
Fig 5.
Peri-ictal radar plots of the fraction of salient values of measure X = {S,N,C,M} in the radiologically confirmed RBT for all 38 seizures in all 16 patients.
Each row shows the data of one outcome subgroup (Engel class I: 6 patients, 14 seizures; class II: 5 patients, 10 seizures; class IV: 5 patients, 14 seizures), across the four different measures (S, absolute EEG slope; N, number of forbidden ordinal patterns; C, surrogate corrected cross-correlation; M, surrogate corrected mutual information). The data is arranged clockwise with the upper right quarter corresponding to three consecutive epochs of one minute duration immediately before seizure onset. The lower half corresponds to the scaled seizure time between seizure onset and termination, and the upper left quarter corresponds to three consecutive one-minute epochs immediately after seizure termination. The red polygons illustrate the temporal profile of the mean and the parametrically estimated 95% confidence interval of the mean is displayed by the thickness of the polygon outline. Broken circular lines correspond to fractions 0.25, 0.5, 0.75 and 1.0 from inside to outside. The continuous circular line corresponds to the fraction f of salient channels expected from the size of the RBT. The mean <F> as estimated from the circle area is given in the lower right corner of each plot. Note that the means given here can deviate from the medians used in the text. Analog figures for the visually defined seizure onset zone (SOZ), its overlap OVL with the RBT and channels contributing to none of these zones NON are given in S5, S6 and S7 Figs.
Fig 6.
Peri-ictal radar plots of the mean for all 38 seizures in all 16 patients.
Note that segment-wise mean log probabilities are displayed, not logarithms of segment-wise p-values. The figure elements are analog to Fig 5 with the exception that dashed circular lines heuristically correspond to p-values of 0.1, 0.01 and 0.001 from inside to outside. The continuous circular line corresponds to p = 0.05. The mean <L> as estimated from the circle area is given in the lower right corner of each plot. Note that the means given here can deviate from the medians used in the text. Analog figures for the visually defined seizure onset zone (SOZ), its overlap OVL with the RBT and the channels contributing to none of these zones NON are given in S8, S9 and S10 Figs.
Table 3.
Testing associations between the Engel class and the seizure-wise mean fraction of salient values in the regions of interest.
Table 4.
Testing associations between the Engel class and the log probability in the regions of interest.