Abstracts

Rationale: VNS (Vagus Nerve Stimulation) Therapy® is an adjunctive therapy for patients with medically refractory epilepsy. The primary metric used to assess response to any treatment for epilepsy is seizure frequency reduction as measured using seizure diaries. Seizure severity reduction is also clinically meaningful to patients and can be measured objectively. Analysis of electro-encephalographic (EEG) signals has revealed that seizures are accompanied by spatial synchronization of EEG electrodes (Neurosci. Lett. 2013; 536; 14-18 and Epilepsy Res. 2015; 113; 98–103). This quantitative feature was obtained from EEG data around ictal events to first evaluate if automated delivery of VNS at the time of seizure onset reduces the severity of seizures in patients by reducing EEG spatial synchronization. We then explored the correlation between the effectiveness of VNS in reducing the severity of seizures and long-term (12 month follow-up) responsiveness using clinical metrics of ≥50% seizure frequency reduction. Methods: 15 patients are included in this study who were experiencing debilitating seizures (complex partial, generalized tonic-clonic, and/or complex partial secondary generalized) and were implanted with the AspireSR® VNS Therapy System. This system automatically delivers stimulation when an increase in heart rate, which may be associated with a seizure, is detected. Baseline seizure data was collected 3 months prior to VNS implantation; follow-up visits were performed 12 months after implant. A total of 103 pre-VNS implant and 102 post-VNS implant seizures were available for analysis. Measure of spatial synchronization of EEG during a seizure is extracted as a feature to quantify the severity of each seizure. The extracted features from all seizures and patients were then provided to an unsupervised Fuzzy C-Means (FCM) clustering algorithm to determine if the selected feature for all patients separates the patients into two distinct classes of response (responders and non-responders). Results: Application of the FCM algorithm resulted in clustering of the patients into two classes of responder and non-responder (Figure 1). The mean and standard deviation of this feature across all seizures and patients are shown in Table 1 along with the corresponding p-value. It appears that the mean value of this feature is significantly lower (p<0.05) for responder patients compared to non-responder patients. Figure 1 demonstrates the classification performance for determining responder and non-responder is 14/15×100%=93.33%. Only patient ‘5’ had a % change in synchronized EEG electrodes that was similar to the non-responder group and was misclassified as a non-responder using the FCM classifier. Conclusions: This is the first study utilized ictal synchronization of brain activities measured with scalp EEG signals as a measure of seizure severity and showed the correlation between decrease in this feature and long term responsiveness to VNS treatment. Based on the correlation with long term response, reduced spatial synchronization may eventually result in the reduction of seizure frequency. Funding: NA