Fig. 1. Time-averaged relative frequency ω_1,1 and mean phase coherence R as a function of coupling strength for the x-coordinates of diffusely coupled Rössler systems. Upper plots: no noise contamination. Lower plots: signals contaminated with additive white noise at an SNR of 1.

Fig. 2. Schematic view of intracranially implanted depth electrodes.

Fig. 3. Typical EEG segments (left) and corresponding normalized power spectra (right). Data were obtained from adjacent recording sites (cf. Fig. 2) during an interictal (no seizure activity), pre-ictal (minutes before an impending seizure), and ictal (during a seizure) state. Computation of the mean phase coherence R (cf. Eq. (8)) for these segments yielded 0.83, 0.37, and 0.94, respectively.

Fig. 4. Time-averaged phase coherence matrices extracted from an interictal recording (no seizure activity, patient No. 1 in Fig. 5). Each matrix reveals two areas of high synchronization (left-hand side: TL1–TL4 and TL5–TL10; right-hand side: TR1–TR3 and TR4–TR10). In ictal segments (seizure activity, not shown here), seizure activity for this patient was first discernible in TL7–TL9.

Fig. 5. Lateralization plots from interictal EEG recordings for 17 patients comparing left and right mean phase coherence averaged over space and time. Asterisk denotes false lateralization in the sense that the mean phase coherence is higher on the nonfocal side.

Fig. 6. Mean phase coherence calculated from interictal and peri-ictal EEG recordings (patient No. 2 in Fig. 5, channels TL1 and TL2) recorded on different days. Seizures are marked by vertical gray bars. Coherence values R>0.04 can be assumed to be statistically significant at a level of p=0.01 (Rayleigh test for uniformity, cf. [26]).

Fig. 7. Same as Fig. 6 but for patient No. 13 in Fig. 5 (channels TL7 and TL8).