Abstracts

Rationale: Responsive neurostimulation may be a useful treatment for epilepsy, and may act in part by modulating ongoing brain rhythms. Here we measured how responsive neurostimulation modulates activity within different frequency bands in patients with epilepsy. Methods: Electrocorticographic (ECOG) signals were recorded from 65 patients participating in a clinical investigation to assess the safety of an implantable responsive neurostimulation system (RNS™ System, NeuroPace, Inc.). In each patient, 4-channel bipolar recordings of one or two epileptogenic regions were made using chronically implanted intracranial electrodes. The cranially implanted neurostimulator processes the signals in real-time, delivers responsive stimulation, and stores ECOG records. Whenever recording sites changed, data from that patient was treated as a new dataset, resulting in 146 datasets. To study rhythmic activity at a frequency, f, we band-pass filtered each recording between f ± 2.5 Hz, then convolved the filtered signal with a wavelet with frequency f to obtain an amplitude and phase. The amplitude measured the strength of activity at frequency f. For each pair of recordings from each dataset, we computed phase differences, converted these to unit vectors in the complex plane, and used the amplitude of the average of these unit vectors to measure the synchrony at frequency f. We compared the strength and synchrony of oscillations 0.2 sec before and 0.8 sec after stimulation, and omitted data containing stimulation artifacts. We measured the statistical significance of phase-locking by bootstrapping. Results: For each frequency and pair of recordings, statistically significant phase-locking occurred in 13% of cases, using a criterion of p < 0.01. After stimulation, significant phase-locking decreased by an average of 19%. Phase-locking decreased for frequencies between 10-80 Hz (p < 0.001), but not between 80-100 Hz. Note that when recordings contained events that in other cases would have led to stimulation, but the stimulator was turned off, we observed marginal and inconsistent changes in phase-locking. When we analyzed all data, we found that following stimulation, rhythmic amplitudes increased between 30-100 Hz (p < 0.01). By contrast, when we only analyzed cases in which statistically significant phase-locking was present, rhythmic amplitudes from 10-80 Hz decreased after stimulation (p < 0.001). Again, these results were not present when recordings contained events that otherwise would have led to stimulation, but stimulation was not delivered. Conclusions: These results suggest that responsive neurostimulation suppresses 10-80 Hz oscillations that are phase-locked across recording sites, but enhances 30-100 Hz oscillations that are not phase-locked across recording sites. Thus, a potential therapeutic effect of neurostimulation may be to suppress large-scale, lower-frequency oscillations while enhancing high-frequency, local oscillations, and selecting stimulation parameters which maximize these effects may help to optimize the therapeutic efficacy of neurostimulation for epilepsy and other disorders.