Bicoherence of Intracranial EEG: A Novel Precursor of Seizure Activity in Canine Epilepsy

• Published in: 2018 IEEE Life Sciences Conference (LSC) pp. 93 - 96
• Main Authors: Gagliano, Laura, Assi, Elie Bou, Sawan, Mohamad, Nguyen, Dang K.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2018
• Subjects: Analysis of variance; bicoherence; Dogs; Electroencephalography; Epilepsy; Feature extraction; Forecasting; Frequency measurement; higher order spectra; preictal period; Seizure forecasting
• DOI: 10.1109/LSC.2018.8572206

Seizure forecasting could improve the quality of life of patients with refractory epilepsy. Seizure prediction relies on the adequate identification of seizure activity precursors from electroencephalography (EEG). However, no single feature is currently capable of characterizing brain dynamics prior to seizures. This work evaluates the suitability of the bicoherence, a measure of non-linear phase-phase coupling interactions based on the higher order spectra (HOS), for seizure prediction in canine epilepsy. Quantitative features were extracted from the bicoherence of 30-sec segments and 2 statistical tests assessed the existence of significant differences between preictal and interictal HOS features. Results of ANOVA tests displayed a statistically significant change during the preictal period as tracked by quantitative bicoherence-extracted measures. Interestingly, over 50% of seizures displayed a significant HOS change in both hemispheres for all 3 dogs diagnosed with focal epilepsy which suggests the possibility that preictal activity may be detectable in remote regions distanced from the seizure onset zone. This work demonstrates the existence of a preictal period characterized by a statistically significant change in bicoherence, highlighting the feasibility of seizure forecasting based on HOS features.