Abstracts

Rationale: Whether epilepsy becomes more intractable with time is a matter of considerable interest. Long-term studies of seizure dynamics would help address this question. Intracranial monitoring is used in the presurgical evaluation of patients with intractable epilepsy, but typically recordings are for only 7 to 10 days. The Responsive Neurostimulation System (RNS™, Neuropace) is an implantable device designed to detect seizure activity and deliver responsive stimulation. Controlled trials are currently in progress. Data obtained from these safety and tolerability trials allow us to look at long term seizure dynamics.Methods: The RNS data of 2 patients who had intracranial cortical strips recording for 21 months were studied. The two patients selected had not had a significant reduction in seizures with the RNS. The data consisted of recordings of events which triggered the RNS device. Only 2 to 4 channels are recorded by the device. From the events recorded by the RNS device, seizures (as determined by visual analysis of EEG activity) were marked. Time-frequency analysis was done by matching pursuit (MP) decomposition. Gabor atom density (GAD) is a signal complexity measure derived from MP (Jouny et al. 2003). Quantitative analysis of the GAD and time-frequency decomposition were done by a visual marking of specific frequency patterns that were present in a majority of the seizures for each patient. The maximum GAD, the duration of the pattern, the maximum frequency of the characteristic pattern (F_max), and the frequency at the end of the pattern (F_end) were determined for each seizure and compared over time.Results: For the first patient, the time-frequency reconstruction of the seizures displayed a clearly identifiable frequency pattern that remained consistent throughout the recording period. The average F_max over 281 seizures was 63±3Hz, and the average F_end was 36±2Hz. The pattern lasted 12.3±1.5s. The frequency pattern remained consistent even after the duration of electrical stimulation from the RNS was increased in the 15th month. GAD_max averaged 0.31±0.02 and decreased to 0.26±0.04 after the change in stimulation parameters. In the second patient, 39 events recorded by the RNS exhibited a consistent frequency pattern. Averaged over these 39 events, F_max was 26±4Hz, F_end was 14±2Hz, and GAD_max was 0.26±0.03.Conclusions: The analysis of the seizure dynamics from these two patients did not reveal any observable changes in patterns in long-term recordings of seizures. These two patients had not obtained benefit from the RNS (possibly due to electrode locations not close enough to seizure foci) but their seizures also did not worsen during the analysis period. Patients whose seizures are reduced by responsive neurostimulation or those patients whose seizures worsen over time may have changing seizure dynamics. These data provide further support for a consistent pattern of seizure dynamics from selected patients, now providing some of the first long-term analyses (almost 2 years). Supported by NIH grant NS 48222, The W. Wilson Undergrad Rsh Fellowship, and Neuropace