Two seizure-free nights with at least two complete sleep cycles were chosen from all subjects. Each night was scored for sleep stages using the software Somnologica™ Studio (Embla Systems, Inc, CO, USA) and visually confirmed by a time-frequency analysis.
Slow waves were first detected from a single scalp electrode and only from segments during NREM sleep. Since not all subjects shared exactly the same number of scalp electrodes, FP1 was chosen for the analysis because it was systematically recorded. For the automatic detection of the slow waves, the signal was first down-sampled to 40 Hz and then bandpass-filtered between 0.1–4 Hz. The criteria for the detection of the slow waves were similar to those used in  and : (1) a negative wave between two succeeding zero-crossings separated by 0.125–1 sec and presenting only one main peak ≤−80 µV and, optionally, other negative peaks not exceeding 50% of the main one (in absolute value); (2) a subsequent (or antecedent) positive wave between two succeeding zero-crossings separated by 0.125–1 sec; (3) a negative-to-positive (or positive-to-negative) peak-to-peak amplitude ≥140 µV (see Figure 1). In order to select only large-extended slow waves, an additional criterion was used to confirm the simultaneous presence of the slow components on intracranial contacts. According to that, we only kept waves for which the average of all intracranial negative peaks (when presented in at least 30% of the total number of contacts) during the scalp positive peak was ≤−60 µV. Finally, we rejected slow waves presenting interictal epileptic discharges inside a window of 2 seconds around the minimal scalp negative peak.
Automatic detection of gamma oscillations
An automatic detection of high-frequency events was performed separately for each intracranial contact and for the whole sleep-wake cycle and independently of the presence of slow waves. In order to detect oscillations, the whole gamma range was subdivided in consecutive sub-bands of 20-Hz each (f2−f1 = 20 Hz, f2>f1), from 30 to 120 Hz. For each one of these sub-bands, oscillations were detected according to a similar procedure than those presented in : First, the envelope of the bandpass-filtered signal was obtained via the Hilbert transform. Then, a threshold for detection was defined consisting on all successive envelope values with amplitudes >3σ above the mean amplitude of the envelope signal (the mean and the standard deviation (σ) computed from free-of-artifacts periods, selected along the whole night by visual inspection). From all detected events, we selected only events having more than 6 oscillatory cycles (calculated from the mean frequency (f1+f2)/2 of the current frequency band), and presenting more than 5 local maxima in both original and filtered signals.
72×10 Isochronic Tones-Alpha, enhanced release of serotonin and mood elevator, universally.