Seizure detection based on vagus nerve activity translation to an implantable setting

Michel-Antony Ngan Yamb Chavez-Cerda, Javier [UCL] Vande Perre, Louis [UCL] El Tahry, Riëm [UCL] Nonclercq, Antoine [UCL]

Epilepsy is a chronic disease that affects fifty million people worldwide. One-third of patients are not responding to antiepileptic drugs. For those patients,if surgery is impossible, vagus nerve stimulation (VNS) can be offered as an adjunctive treatment. As on-demand stimulation improves VNS efficacy, closed-loop VNS – i.e., stimulating at the occurrence of a seizure – emerges as a promising solution for those patients. However, a suitable biomarker for seizure detection is required. Besides, to be clinically valuable,computations related to such a biomarker should be low enough to be embedded in an implant. This work explores the use of vagus nerve electroneurogram (VENG) as a biomarker and compares the performances of various seizure detection algorithms. We analyzed data from six rats that exhibited acute seizures induced by pentylenetetrazol (PTZ), along with three control rats infused with a saline solution. All seizures could be detected with no false positives, even without requiring large computations. This work opens ways to reduce the computation power necessary for VENG-based seizure detection, aiming at embedding a seizure detection algorithm in an implant. Further work should address whether the performance of low computation power algorithms is maintained in a chronic setting, i.e., in the presence of movement artifacts.