Treatment of an Epilepsy-Associated Disorder

Abstract

The present invention relates to an apparatus for the treatment of an epilepsy-associated disorder in a living being and to a method for the treatment of an epilepsy-associated disorder in a living being.

Claims

1. An apparatus for the treatment of an epilepsy-associated disorder in a living being, comprising, in an operatively linked manner: an electrode set capable of obtaining an electroencephalogram (EEG) of a living being; an audio output device capable of delivering an audio signal to the living being; an analysis and control unit capable of: (i) detecting interictal epileptic discharges (IED) in the EEG of the living being, and (ii) causing the audio output device to deliver the audio signal in time dependence of a detected IED.

2. The apparatus of claim 1, wherein the analysis and control unit is configured to cause the audio output device to deliver the audio signal to the living being at a time delay after a detected IED.

3. The apparatus of claim 2, wherein the time delay is between approx. 1 to 3 seconds relative to the negative peak of the detected IED.

4. The apparatus of claim 1, wherein the audio signal comprises 1/f noise.

5. The apparatus of claim 1, wherein the analysis and control unit is configured to cause the audio output device to deliver the audio signal to the living being for a time period of approx. 50 ms.

6. The apparatus of claim 1, wherein the analysis and control unit is configured to cause the audio output device to deliver the audio signal to the living being with a volume of approx. 12 dB above the individual hearing threshold of the living being.

7. The apparatus of claim 1, wherein the analysis and control unit comprises a band-pass filter for detecting the IED in the EEG of the living being.

8. The apparatus of claim 7, wherein the band-pass filter comprises frequency limits of approx. 20 and approx. 150 Hz.

9. The apparatus of claim 1, wherein the analysis and control unit is configured to run a threshold method using an amplitude threshold for detecting the IED.

10. The apparatus of claim 1, wherein the electrode set comprises a detection electrode or a reference electrode.

11. The apparatus of claim 1, wherein the analysis and control unit is further capable of: (iii) determining the non-rapid eye movement (NonREM) sleep phase of the living being.

12. The apparatus of claim 11, wherein the analysis and control unit capable of: (ia) detecting IEDs in the EEG of the living being during a NonREM sleep phase, and (iia) causing the audio output device to deliver the audio signal in time dependence of a detected IED during a NonREM sleep phase.

13. The apparatus of claim 1, wherein the audio output device is a headphone.

14. The apparatus of claim 1, wherein the living being is or is suspected from suffering from any of: epilepsy, childhood epilepsy or Rolandic epilepsy.

15. A method for the treatment of an epilepsy-associated disorder in a living being, comprising the following steps: (1) obtaining an electroencephalogram (EEG) of a living being; (2) detecting interictal epileptiform discharges (IED) in the EEG of the living being, and (3) administering an audio signal to the living being in time dependence of a detected IED.

16. The method of claim 15, wherein the audio signal is administered to the living being at a time delay after a detected IED.

17. The method of claim 16, wherein the time delay is between approx. 1 to 3 seconds relative to the negative peak of the detected IED.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0068] FIG. 1 Schematic illustration of an embodiment of the apparatus according to the invention.

[0069] FIG. 2 Stimulation protocol. In a study assessing the efficacy of the invention, epileptic spikes (IEDs) were detected using an amplitude threshold. Auditory stimulation was performed either at the time of an IED's negative peak (“Negative peak” condition), its positive peak (approximated by introducing a stimulation delay of 90 ms, “Positive peak” condition), 0.5 s after the negative peak (“0.5 s delay” condition), or at a random delay between 1 and 3 s after the negative peak (“Random delay” condition). In a “Sham” control condition, no stimulus was delivered upon a detected IED. In a healthy control group, only the Random delay and Sham conditions were executed. Stimulation time points were selected randomly during NonREM sleep and with the same temporal structure as in the patient group. The stimulation conditions were conducted in blocks of 30 s.

[0070] FIG. 3 Spike detection and EEG signal interpolation. From the original timeseries data (first trace from top) the inventors extracted components showing prominent IEDs (second trace, head plot shows highly localized spike topography). After highpass-filtering, a threshold (straight horizontal line) was determined and used to detect spikes (third trace). Detected IEDs were removed and the EEG signal was interpolated, resulting in an IED-free signal (fourth trace). Vertical grey bars indicate the detected IED events.

[0071] FIG. 4 Acoustic random-delay stimulation suppressed IEDs in all patients. A) IED density was modulated by the stimulation (ANOVA on pooled data, p=0.034) and significantly lower in the Random delay condition compared to Sham (p=0.015, Holm corrected). B) IEDs were suppressed in each of the seven patients (paired t-test on individual data, p=0.020). C) Representative examples for 30-s blocks without stimulation (Sham, 26 detected IEDs) and Random delay stimulation (Random delay, 18 detected IEDs). Triangles over dark lines show sham and random delay stimulations, respectively. Grey bars mark detected IEDs. Note that the second stimulation in the Random delay condition (at around t=14 s) by chance occurred close to another IED. These coincidences present a negative bias, suggesting that the results underestimate the actual effect of the stimulation. *p<0.05.

[0072] FIG. 5 Rolandic epilepsy is associated with slower sleep spindles. Results of analysis of IRASA-derived power estimates and spindle detection. (A) Oscillatory component of IRASA power spectra relative to fractal 1/f component. Note clear peaks in the spectrum for slow wave, theta, and spindle frequency bands. (B) Comparison of slow wave power, theta power, spindle power, spindle peak frequency, and spindle rate between patients and controls. Sleep spindles reached their peak power at a lower frequency in patients than in controls (p=0.002). All data shown as mean±SEM for sham condition and electrode Cz. **p<0.01.

[0073] FIG. 6 IED rate is negatively correlated with spindle rate and power. Across all conditions, higher spike rates were associated with lower spindle power and rate, pointing to a competitive relationship between epileptic activity and sleep spindles. Each grey circle represents a single 30-s stimulation block, the regression line is shown in black ±bootstrapped 95% confidence interval (both Pearson r<−0.231 and p<2.019×10.sup.−17). Rates are integers leading to largely overlapping data points. To increase clarity, values are slightly jittered along the y axis and marginal histograms illustrate the distribution of values across each axis.

[0074] FIG. 7 Neural responses to IEDs and tones. A) Neuronal responses to IED-triggered acoustic stimulation in the patient group, aligned to detected IEDs (or tone in Random delay condition) at time 0. A) Each panel shows the time-frequency representation (top; colors denote power change relative to baseline −1.5 to −0.5 s before the IED negative peak) and time-domain representation (bottom; bands show 95% confidence interval) of the response. Responses in the Negative peak, Positive peak, and 0.5 s delay conditions were compared to Sham (power: black contours show significant clusters, non-significant areas are covered by a half-transparent mask; time-domain: horizontal bars show significant clusters, all cluster level p<0.05). B) Evoked spindle power (10-18 Hz, 0.4-1.2 s after IED negative peak, ±boot-strapped SEM), relative to baseline, extracted from time-frequency data in A). In patients (light grey), auditory stimulation triggered spindle responses of similar strength as those evoked by epileptic spikes (post-hoc test Random delay vs. Sham, Holm-corrected p=1.00). In contrast, in control subjects (dark grey), the Sham condition did not contain stimulation or spikes, leading to spindle power around baseline levels and significantly lower than for Random delay stimulation (Random delay vs. Sham, Holm-corrected p=0.033; ANOVA Patient/Control x Sham/Random delay, p=0.019). Stimulation-induced spindle power did not significantly differ across groups (Random delay patients vs. controls, Holm-corrected p=1.00). C) Evoked spindle power differed significantly across the stimulation conditions (light grey line, ANOVA Sham/Negative peak/Positive peak/0.5 s delay, main effect Condition, p=0.007). When subtracting the isolated auditory-evoked response of the Random delay condition (stimulation-aligned) before assessing spindle power, this difference vanished (dark grey, p=0.752).

[0075] FIG. 8 Tones decreased the amplitude of immediately following IEDs. Amplitude of spikes following auditory stimulation or sham (0-1.5 s) were reduced in amplitude (ANOVA on pooled data, main effect Condition p=0.002). When compared to Sham, this reduction was significant for the Positive delay condition (Holm-corrected p=0.021; 0.5 s delay, p=0.121; Random delay, p=0.250). *p<0.05; **p<0.01.

EXAMPLES

1. Embodiment of the Apparatus

[0076] In FIG. 1 a schematic illustration of an embodiment of the apparatus according to the invention is depicted, said apparatus is shown under the reference sign 10. The treated living being or child has the reference sign 12. The apparatus 10 comprises an electrode set 14 attached to the head of the child 12. The child 12 wears headphones 16 and, in the illustrated embodiment, is in a sleeping state. The electrode set 14 and the headphones 16 are connected via cables with an analysis and control unit 18. The analysis and control unit 18 is capable of detecting interictal epileptic discharges (IEDs 20, marked by grey vertical bars) in the EEG 22 (oscillating line; vertical straight line shows detection threshold) of the child 12. The analysis and control unit 18 is delivering an audio signal in time dependence of a detected IED 20 to the child 12 via the audio output device 16.

[0077] The electrode set 14, in the illustrated embodiment, comprises a detection electrode 24, a reference electrode 26, and, optimally, additional electrodes 28 for a better determination of the sleep phase of the child 12.

2. Introduction

[0078] In the illustrated study the inventors aimed at lowering IED rates during sleep in children with Rolandic epilepsy using auditory stimulation. The short tone bursts presented during sleep were expected to suppress spikes either by the immediate induction of a SO-like down-state or by inducing up-state related thalamic spindles interfering with thalamocortical IEDs generation. To explore these hypotheses, the inventors compared the effects of auditory stimulation between different conditions with the tone presented either immediately upon detection of a spike (“Negative peak” condition), 90 ms later (“Positive peak”), 500 ms later (“0.5 s delay”), or in a random delay interval 1-3 s following the negative peak of the spike (Random delay”). Effects on spike rates were compared with a sham control condition, and the inventors also compared the effects of sham and random stimulation in patients with those in healthy age-matched controls.

3. Material and Methods

Participants

[0079] A total of 14 subjects (7 females, 7 males; mean age ±SD: 9.97±1.52; range: 6.60-11.76 years) participated in the study. Half of the participants (“patients”) had previously been diagnosed with Rolandic epilepsy, benign epilepsy with centro-temporal spikes (BECTS), or BECTS-typical centro-temporal spikes, and did not show any known structural neuronal abnormalities. The other participants (“controls”) had not been diagnosed with epilepsy or any other known neural disorders. Controls were matched to the patients by age (patients: 9.86±1.65 years; controls: 10.08±1.50 years; p=0.798) because the investigated form of epilepsy as well as physiological sleep characteristics change substantially with age. Four of the patients were on epilepsy medication. See Table 1 for further details. Two additional participants were excluded from analysis due to technical problems during the experimental night. Further four subjects were excluded who had previously presented with BECTS-typical centro-temporal spikes but showed no or almost no epileptic activity during the experimental night.

TABLE-US-00001 TABLE 1 Participant details. Detection Subject Sex Age electrode Medication Patients 1 m 9.3 T4 Oxcarbazepin (2 × 450 mg) 2 w 10.8 T4 Levetiracetam (2 × 500 mg) 3 w 10.5 T3 4 m 6.6 Cz 5 m 11.8 T4 Ospolot (2 × 50 mg) 6 w 9.6 C3 Ospolot (2 × 50 mg) 7 w 10.5 C3 Controls 8 m 6.9 C3 9 m 9.6 C3 10 m 10.9 Cz 11 w 10.8 C3 12 m 11.4 F4 13 w 10.2 Cz 14 w 10.8 Cz

[0080] Participants in both groups were matched by age (no difference between groups, p=0.798). The detection electrode was chosen based on previously determined epileptic focus and spike amplitude in the recording night. Medication dose in Table 1 is given per day.

[0081] Table 2: Potential patients were preselected, approached, and informed about the study by an experienced pediatrician (S.R.) with access to the candidate's medical history. Potential control participants were contacted via the institute's volunteer database. Exclusion criteria were other neurological or psychological disorders, irregular sleeping patterns, or ongoing participation in other studies. The participants' parents gave their written informed consent, subjects gave their verbal consent, and both were free to abort the study at any stage. The study was approved by the local ethics committee of the Medical Faculty of the University Tubingen.

TABLE-US-00002 TABLE 2 Sleep parameters. Subject # S1 S2 S3 S4 REM TST Patients 1 15.0 204.0 178.5 60.5 40.5 505.0 2 68.0 390.0 66.0 20.0 95.0 643.0 3 37.0 214.0 126.0 76.0 87.5 572.0 4  1.5  47.5 461.0 98.5 16.0 639.0  5* 24.0 103.5 28.5 57.0 27.0 250.5 6 20.0 162.5 254.0 53.5  0.0 496.0 7 23.5 259.0 112.5 57.0 95.0 553.5 Mean 27.50 ± 9.37 212.83 ± 46.00 199.67 ± 58.48 60.92 ± 10.62  55.6 ± 17.33 568.08 ± 25.85 Controls 8 15.0 151.0 128.0 40   69.0 404.0 9 59.0 220.0 77.0 65.5 89.5 526.0 10  51.0 267.5 94.0 70.5 93.0 631.5 11   8.5 155.0 184.0 101.5  96.5 569.5 12  65.0 151.0 144.5 55.0 50.5 583.0 13  31.5 135.0 89.5 69.0 25.5 419.0 14  26.5 280.5 69.5 63.5 78.5 553.5 Mean 36.34 ± 8.31 194.29 ± 23.02 112.36 ± 15.70 66.43 ± 7.06  71.79 ± 9.79 526.64 ± 32.13

[0082] Time spent in each sleep stage (in min). There were no differences between groups in any sleep stage (independent t-tests, all p<0.15 uncorrected, subject 5 excluded from means and statistics). *For participant 5, polysomnographic recordings were only available for the first half of the night. Because the patient had difficulties falling asleep after a bathroom break the EEG was removed; the patient's mother reported normal sleep in the second half of the night.

Experimental Procedure

[0083] Accompanied by one parent, participants arrived at the laboratory at around 7:00 pm. EEG and other polysomnographic electrodes were attached, earphones were put on, the individual hearing threshold was determined (43.00±2.13 dB sound pressure level, mean±SEM), and sleepiness was rated verbally on the Stanford Sleepiness Scale. At around 10:00 pm, participants went to bed. Parents either slept in the same room as the child or in a room next door. Acoustic stimulation was started as soon as the polysomnography indicated reliable sleep patterns and continued for about 3 h (174.14±13.94 min, mean±SEM). Depending on the participant's sleep pattern and frequency of detectable spikes, this resulted in 183.19±18.48 stimulations per subject and condition (see below). After the stimulation period, participants continued sleeping without further interventions.

Electrophysiological Recordings

[0084] Throughout the night, the inventors acquired electroencephalographic recordings at 19 electrode sites (Fp1, Fp2, F3, Fz, F4, F7, F8, C3, Cz, C4, T3, T4, T5, T6, P3, Pz, P4, 01, and 02, according to the International 10-20 system), referenced to the mastoids (averaged M1, M2). Additionally, the inventors obtained bipolar electromyographic recordings from the chin as well as horizontal and vertical electrooculography to aid sleep scoring. Data were recorded and amplified with a Brain Products recording system (Brain Products GmbH, Gilching, Germany) at a sampling frequency of 500 Hz and processed using Matlab 2017a (Mathworks, Natick, USA), Fieldtrip (fieldtriptoolbox.org), Python 3.7, Seaborn 0.9.0 (seaborn.pydata.org). In most cases, analysis was restricted to the Cz electrode site as well as the individual detection electrode (DET) nearest to the focus of the patient's spike activity. For control subjects, the same electrode as the age-matched patient was chosen as detection electrode.

Auditory Stimulation

[0085] The EEG electrode closest to the known epileptic focus was used to detect epileptic spikes. The signal from this detection electrode (DET) was passed on to the stimulation setup, referenced to the EEG system's mastoid electrodes. In cases in which recordings showed higher spike amplitudes at another electrode, the detection electrode was switched before starting the stimulation. Activity at the detection electrode was recorded by a Digitimer D360 EEG amplifier (Digitimer Ltd., Welwyn Garden City, UK), using the same reference as the EEG system. The signal was real time-filtered between 4 and 150 Hz. The DET signal was sampled at 200 Hz by a CED Power1401 MK2 data acquisition interface (Cambridge Electronic Design, Cambridge, UK). Spike detection was performed via a custom-made script running under Spike2 7 (Cambridge Electronic Design, Cambridge, UK) using a threshold procedure. The script triggered acoustic stimulation by short bursts of pink noise (50 ms duration, +12 dB above previously established hearing threshold), which were administered using in-ear headphones. Detection threshold and frequency range for the online filtering were continuously adapted to maximize spike detections (true positives) while minimizing false detections of other events (false positives).

[0086] The inventors performed auditory stimulation during Non-REM sleep (stages 2, 3, and 4) in five conditions, randomly selected for blocks of 30 s (FIG. 2). Tones were presented either at the time of the negative peak of a detected spike (“Negative peak”), or 90 ms later during its subsequent positive peak (“Positive peak”), or 0.5 s after the negative peak (“0.5 s delay” condition), or with a random delay between 1 and 3 s after the negative peak (“Random delay”). In a “Sham” control condition, the negative peak of a spike was detected but no auditory stimulation was performed. Stimulation was paused at signs for arousal, awakening, or REM sleep.

Data Analyses

[0087] IED detection and removal by EEG signal interpolation. The inventors detected interictal epileptiform discharges automatically using custom-made algorithms. The inventors first performed an Independent Component Analysis (ICA) decomposition (fastICA) and manually selected ICA components that did not contain IED-like waveforms for rejection. After a back-projection into EEG space, the inventors highpass-filtered the EEG signal of the detection electrode at 5 Hz, extracted the envelope using a Hilbert transform, and smoothed the envelope with a moving average of 50 ms length. For each subject, the inventors determined the detection threshold as the mean+2.5 times the standard deviation (SD) of the envelope signal over all NonREM periods. For one subject, the scaling factor was adjusted to 2.0. An IED was registered whenever the enveloped exceeded the threshold for more than 10 and less than 500 ms. Putative events following within <50 ms were merged. For each IED event, the inventors marked on- and offset where the envelope crossed the threshold. The inventors then calculated the IED rate, i.e. events per minute, separately for each experimental condition. To assess differences in IED rate between condition, the inventors pooled IED rates for all 30-s stimulation blocks across patients and then performed an Analysis of Variance (ANOVA) with a fixed Factor ‘Condition’ (Negative peak/Positive peak/0.5 s delay/Random delay/Sham) and IED rate as the dependent variable. Post-hoc t-tests were subjected to Holm correction for multiple comparisons. To remove confounding effects of IEDs in subsequent analyses, the inventors created an IED-free dataset by performing a spline interpolation on the EEG time-domain signal for each identified IED with additional padding of 100 ms on each side (FIG. 3).

[0088] Spectral analysis. To assess spectral power (for patients on the interpolated signal), the inventors segmented data for each experimental condition into epochs of 4.096 s with an overlap of 50%. The inventors estimated the frequency content of the signal using the irregular-resampling auto-spectral analysis (IRASA) approach. This procedure allows the isolation of oscillatory spectral components from fractal 1/f background component typical for electrophysiological recordings. The inventors expressed the magnitude of the oscillatory component relative to the fractal component. Power was assessed by averaging the resulting values for the slow wave (0.5-4 Hz), theta (4-8 Hz) and spindle (10-18 Hz) bands, respectively. Peak spindle frequencies were determined individually for each subject. The inventors assessed statistical differences in theta and spindle power as well as spindle peak frequencies by performing a repeated measures ANOVA with between-subject factor ‘Group’ (Patients/Controls) and within-subject factors ‘Electrode’ (Cz/DET) and ‘Condition’ (Random delay/Sham).

[0089] Offline detection of SOs and sleep spindles. Discrete slow oscillation and spindle events were detected during NonREM epochs across the entire night. For SOs, each EEG channel was first bandpass-filtered from 0.3 to 1.25 Hz. Then positive to negative zero crossings were identified and all intervals between consecutive zero crossings shorter than 0.8 or longer than 2 s (corresponding to frequencies of 0.5-1.25 Hz in the SO range) were discarded. Across the remaining intervals the negative peaks and the amplitude from negative to the positive peak were averaged. The resulting mean values were multiplied by 1.5 and served as detection threshold, i.e. intervals were labelled as a SO whenever its negative peak was lower than 1.5 times the mean negative peak value and the amplitude exceed the 1.5 times the mean amplitude threshold. To detect sleep spindles, the inventors bandpass-filtered the EEG data between 10 and 18 Hz and derived the root-mean-square (RMS) of the signal with a 200-ms sliding window, followed by smoothing with an identical window length. The inventors determined a detection threshold from the mean RMS signal across all NonREM epochs+1.5 times its SD. Based on this threshold, intervals for which the RMS signal exceeded the threshold for more than 0.4 s and less than 3 s were labelled as discrete spindle events. The inventors determined SO and spindle rate, i.e., the number of detected events per minute, separately for each experimental condition. Statistical assessment of SO and spindle density was based on a repeated measures ANOVA with the between-subject factor ‘Group’ (Patients/Controls) and the within-subject factors ‘Electrode’ (Cz/DET) and ‘Condition’ (Random delay/Sham). The inventors further examined how sleep spindles relate to IEDs by estimating the rate, i.e. events per minutes, of detected spindles and spindle power (10-16 Hz) for all 30-s stimulation blocks. The inventors correlated spindle rates and IED rates (for the same blocks) across all patients.

[0090] Analysis of evoked responses. For analyses of IED-related and stimulation-related activity, non-interpolated data were cut into segments of 14 s around detected spikes. Segments containing artifacts were rejected using a semi-automatic procedure (identification of candidate artifacts by thresholding the z-transformed signal, followed by thorough visual inspection). Data were highpass-filtered at 0.1 Hz and notch-filtered between 45 and 55 Hz (Butterworth filter). Evoked potentials were corrected by subtracting an average baseline value between −1.5 and −0.5 s before the detected spike. For time-frequency analyses 12-cycle Morlet wavelets were applied and visualized as power change relative to the average power at that frequency between −1.5 and −0.5 s.

[0091] Statistical analyses for evoked potentials and evoked time-frequency response relied on non-parametric cluster-permutation statistics to control inflation of type I errors due to multiple comparisons. For time-frequency data, samples were selected that showed significant differences in power in relation to the respective contrast condition (two-tailed paired-samples t-tests, sample-level alpha=0.05). In the resulting statistical map, adjacent samples were grouped into positive and negative clusters for which cluster-level statistics were calculated by summing up the t-values within each cluster. These were tested against a reference distribution (cluster-level alpha=0.05), generated by shuffling the association of data and condition (10,000 permutations) and, for each permutation, taking the maximum statistics among all clusters.

[0092] In a separate step, the inventors isolated evoked spindle power by averaging time-frequency data over all samples within a spectral-temporal segment after applying a broad Tukey window (α=0.25) to attenuate power at the edges. The analyzed segment corresponded to power occurring in the 10-18 Hz spindle band and 0.4-1.2 s following spike onset (negative peak). The spectral and temporal width of the window was determined based on the grand average over all control subjects for the Cz electrode. The inventors statistically assessed evoked spindle power by performing a repeated measures ANOVA with factors ‘Electrode’ (Cz/DET) and ‘Condition’. Results were Greenhouse-Geisser-corrected if the assumption of sphericity was violated. Effect sizes (n.sup.2 for ANOVA, Cohen's d for post hoc tests) are provided for significant tests.

4. Results

Acoustic Stimulation Reduced Interictal Epileptic Discharges

[0093] To assess the effect of auditory stimulation on the occurrence of IEDs, the inventors automatically identified discrete IED events and examined the spike rate across the different stimulation conditions. IED rate was determined for each stimulation block and pooled across all patients. This analysis revealed a striking reduction in IEDs driven by a reduction in the Random delay condition (23.319±0.682 events/min) in comparison to the Sham condition (26.452±0.764 events/min) corresponding to a decrease of ˜11% (F(4,1328)=2.615 and p=0.034, η2=0.008, with Holm corrected p=0.015 for Random delay vs. Sham post-hoc test).

Patients with Rolandic Epilepsy Express Slower Sleep Spindles

[0094] In light of the distinct effect of Random stimulation decreasing spike rate in comparisons with the Sham control condition, in subsequent analyses the inventors aimed to clarify the mediating mechanism of this effect. Specifically, the inventors wondered whether this effect is linked to alterations in slow wave and spindle activity, or might be a consequence of generally lightened sleep during random stimulation. To this end, as a first step, the inventors compared physiological oscillations between the conditions and between patient and control groups using IRASA-based spectral power analyses of the EEG signal.

[0095] An overall examination of spectral power in Sham and Random delay conditions revealed distinct peaks in the 0.5-4 Hz slow wave, 4-8 Hz theta, and 10-18 Hz spindle bands in both groups (FIG. 5A). Testing across groups (Patient/Control), conditions (Random delay/Sham), and electrodes (Cz/DET) did not result in significant differences in slow wave power (F<2.594, P >0.133) or theta power (F<2.877 and p >0.116 for all main effects and interactions) between stimulation conditions or groups (FIG. 5B). Power in the spindle band was as expected higher at Cz than the detection electrode (DET, main effect Electrode, F(1,12)=11.292, p=0.006, η2=0.124), and slightly lower during Sham than Random stimulation (main effect Condition, F(1,12)=4.819, p=0.049, η2=0.004; Electrode x Condition, (F(1,12)=8.760, p=0.012, η2=0.001, FIG. 4D). Intriguingly inspection of the spectra showed that the spindle peak was at a slower frequency in patients than in the controls (patients=11.69±0.15 Hz, controls=13.27±0.22 Hz; main effect Group, F(1,12)=14.53, p=0.002, η2=0.548). Moreover, independent of the group, in the Random delay condition, the spindle peak frequency was slightly faster than in the Sham condition (main effect Condition, F(1,12), p=0.018, η2=0.004, suggesting externally elicited spindles to be faster than spontaneously occurring spindles. Analyses of SO rates revealed more SOs after Random stimulation (F(1,12)=8.882, p=0.011). Spindle rates revealed a higher density at Cz than the detection electrode site (F(1,12)=9.17, p=0.010, η2=0.094, Electrode main effect).

[0096] Overall, the distinct differences in spindle activity between stimulation conditions as well as between the patient and healthy control groups, in the presence of rather comparable slow activity led the inventors to assume sleep spindles to be more relevant than SOs for conveying the suppressing effect of Radom delay stimulation on the patient's spike rates, and the inventors subsequently focused analyses on spindles. Moreover, because slow oscillatory activity is a sensitive measure of sleep depth, these measures being unaffected by Random delay in comparisons with the Sham control condition, excludes that effects of this stimulation spike rates were mediated by a nonspecific lightening of sleep.

Does the Effect of Acoustic Stimulation Result from a Competitive Relationship Between Sleep Spindles and IEDs?

[0097] Considering the partial overlap in the thalamo-cortical circuits mediating sleep spindles and IEDs, the inventors asked whether spindles might lower IEDs rates by occupying these networks and thereby preventing a simultaneous generation of IEDs. To test this, the inventors correlated the rate of IEDs with spindle power and spindle rates within 30-s stimulation blocks pooled across patients. This analysis revealed moderate but highly significant negative correlations for all stimulation conditions (r<−0.231 and p<2.019×10-17 for spindle power and spindle rate correlations across all conditions; all r<−0.135 and p<0.023, uncorrected, for correlations within individual conditions; values taken from detection electrode, FIG. 6) which is consistent with the hypothesis of a principal competitive relationship between pathological epileptic and physiological spindle activity.

Spikes and Tones Evoke Comparable Spindle Activity

[0098] Up to here, the inventors demonstrated an effect of acoustic random delay stimulation on the occurrence rate of IEDs. The inventors now assessed the activity associated with both spikes and tones and whether this activity was modulated by different stimulation conditions. FIG. 7A shows grand averages of time-frequency resolved event-related EEG responses to the IED for the Sham condition, the three peri-spike stimulation conditions, and the Random delay condition in the patient group (data from control group not shown). Visual inspection of IED-evoked changes in spectral power in the Sham condition (top row) shows that spikes are followed by a distinct increase in 10-18 Hz spindle power. The peri-spike stimulation conditions (rows 2-4) show a similar response, with the addition of characteristic auditory evoked power increases. In order to statistically dissociate the response to the tone, the inventors tested the peri-spike stimulation conditions against Sham. This resulted in significant differences with clusters spanning the spindle range and lower event-related frequencies. An increase in a similar frequency band emerged after auditory stimulation in the Random delay condition (bottom row), where spike-evoked activity is approximately averaged out. Overall, these analyses revealed that IEDs and tone invoked comparable increases including and most importantly in the spindle frequency range. Following the inventors' hypothesis of a competitive interaction between spindles and spikes, the inventors more closely compared the spike-evoked (Sham condition) and tone-evoked (Random delay condition) spindle responses for both the patients and control children (power estimates averaged between 10 and 18 Hz and from 0.4 to 1.2 s, FIG. 7B). In this analysis, healthy controls remained at baseline spindle activity in the Sham condition (due to missing spikes in these children) but distinctly increased spindle activity in response to the tones presented in the Random delay condition (Random delay vs. Sham, Holm-corrected p=0.033, d=0.903). By contrast, the patients showed spindle increases in response to both IEDs (Sham) and tones (Random delay) with a comparable magnitude of these responses (Random delay vs. Sham, Holm-corrected p=1.00; p=0.019, eta.sup.2=0.117, for ANOVA Patient/Control x Sham/Random delay interaction.

Do Spike- and Tone-Evoked Responses Interact?

[0099] After demonstrating that IEDs and tones in isolation evoke comparable spindle responses, the inventors explored whether the hypothesized competitive relationship of IEDs and tones is expressed in an interference in their respective evoked activity. The inventors assumed that, when the thalamo-cortical network is occupied with generating a spike and associated spindle activity, auditory stimulation would result only in a dampened spindle response. According to the opposing null hypothesis, activity in the peri-spike stimulation conditions would be a mere superposition of spike- and tone-evoked activity. To discern these two scenarios, the inventors compared the spectral power response to the IED in the Sham condition (reflecting the pure response to the spike in the absence of any tone) with the response to the IED in the Negative peak, Positive peak, and 0.5 s delay stimulation conditions (FIG. 7 C, grey line). Spindle power was significantly different across these four conditions (main effect Condition p=0.007, eta.sup.2=0.052; test performed excluding Random delay condition). As a next step, the inventors subtracted tone-evoked responses from the peri-spike stimulation conditions before extracting spindle power. For estimating the tone-evoked power responses, the inventors used the Random delay condition because in this condition, IED-evoked modulations are assumed to average out to zero, resulting in a pure auditory response. FIG. 7C (red line) shows that after subtracting tone-evoked power changes, spindle responses in the Negative peak, Positive peak and 0.5 s delay conditions do not differ anymore from those in the Sham condition (main effect Condition, p=0.752). These results suggest that spindle responses in the peri-spike stimulation conditions are the result of a mere superposition of responses evoked by spikes or tones in isolation, without strong mutual interactions. Therefore, the competitive effect of tone-evoked activity and IEDs does not extend to IED-associated spindle activity, making a contribution of the latter to IED suppression unlikely.

[0100] A possible immediate interaction between tone- and IED-related activity may also manifest in the other direction, i.e., an effect of tones on the morphology of immediately following IEDs. Therefore, in a complementary analysis, the inventors focused on spikes that occurred within a 1.5-s interval following a tone stimulation (FIG. 8). Compared with respective spikes in the Sham condition, spike amplitudes in the 1.5-s post-stimulation interval were significantly altered (F(4,4573)=4.274, p=0.002, n.sup.2=0.004). Post-hoc tests against Sham showed descriptive reductions in all except the Negative peak condition and reached significance for the Positive peak condition (Holm corrected p=0.021, d=−0.138). In summary, while the results do not demonstrate an immediate effect of spikes on the expression of tone-evoked activity, they show that tones reduced the amplitude of subsequent IEDs, if the stimulation is not administered right at its negative peak.

5. Discussion

[0101] The inventors show by experiments that auditory stimulation during NonREM sleep in children with Rolandic epilepsy or BECTS does suppress pathological interictal neural activity when timed mainly outside the close temporal proximity of IEDs. Across conditions, rate and power of spindles were negatively correlated with the rate of IEDs. The inventors hypothesize that this effect results from a stimulation-induced spindle refractoriness affecting subsequent IEDs that would have otherwise emerged from the same thalamocortical networks. Importantly, analysis of slow wave activity and slow oscillation events confirms that the effect of stimulation is not merely the result of lightened sleep. On the contrary, SO density was even slightly increased in the most effective Random delay condition, suggesting deeper sleep.

[0102] What could be the neurophysiological origin of a competitive relationship between sleep spindles and IEDs? Auditory stimulation induces cortical down states, which in turn lead to thalamic hyperpolarization, associated rebound bursting in thalamo-cortical cells, and thereby sleep spindles. The inventors propose a close relationship between these spindles and pathological IEDs as previous investigations have shown large overlaps in their thalamocortical origin and circuit-level mechanisms. The inventors' results extend these findings by showing that the peak spindle frequency is lower in patients than healthy controls, which may suggest that IEDs lead to pathological alterations of the same thalamocortical system that also generates spindles. Furthermore, sleep spindles show a refractory period of up to several seconds, resulting from intra-thalamic changes in excitability and refractory periods have also been shown for interictal spikes. Refractory periods may thus be shared by the two phenomena, such that spindles inhibit the occurrence of subsequent IEDs. This hypothesis would explain why IEDs and spindles are negatively correlated across conditions, while stimulation only suppressed IEDs when delivered mostly outside of epileptic discharges. Once an IED has triggered a thalamocortical refractory period, eliciting a spindle-associated refractory period in additional, spike-unaffected thalamocortical connections would remain without effect on the generation of further IEDs. However, when tones are applied mostly outside of ongoing epileptic discharges, they may capitalize on the negative spindle-IED relationship and suppress further discharges.

[0103] Similar to tones, also IEDs induce thalamic hyperpolarization, subsequent bursting rebound, and associated spindle oscillations. In the inventors' analyses, the spectral composition of these IED-induced spindles did not significantly differ from stimulation-induced spindles. The inventors further investigated whether an interaction of spindles triggered by IEDs and those triggered by tones could provide further insight into the origin of the demonstrated suppression effect. However, the inventors' analyses did not show a significant interaction of IED- and tone-induced activity. Subtracting auditory-evoked activity from the electrophysiological signal resulted in spindle levels comparable across sham and peri-spike stimulation conditions. IEDs as rather local phenomena may not affect global auditory processing to an extent discernible by the current study. Source reconstructing high-resolution electrophysiological recordings may allow investigating the respective effects within networks comparably affected by tone- and IED-induced spindles.

6. Conclusion

[0104] To summarize, the inventors present evidence for a therapeutic effect of auditory stimulation on interictal epileptiform discharges epilepsy patient, in particular in children with Rolandic epilepsy or RE-like centrotemporal spikes. As the underlying mechanism, the inventors propose a competitive mechanism of sleep spindles and IEDs, based on commonalities in their thalamocortical generation. Already 30 years ago, Mircea Steriade proposed that “[ . . . ] any pharmacological manipulation that would decrease or abolish the inhibitory efficacy of the [reticular] spindle pacemaker upon thalamo-cortical neurons [ . . . ] would also decrease the incidence of epileptic [spike wave] discharges” (Steriade, 1990). With this study, the inventors present first evidence that, as an alternative to pharmacological interventions, entirely non-invasive acoustic manipulation does have an equivalent effect.