Modulated high frequency oscillations can identify regions of interest in human iEEG using hidden Markov models

Author

Guirgis, Mirna ; Chinvarun, Yotin ; del Campo, Martin ; Carlen, Peter L. ; Bardakjian, Berj L.

Author_Institution

Inst. of Biomater. & Biomed. Eng. (IBBME), Univ. of Toronto, Toronto, ON, Canada

fYear

2015

fDate

22-24 April 2015

Firstpage

926

Lastpage

929

Abstract

This study investigated the seizure and non-seizure state transitions in the intracranial electroencephalogram (iEEG) recordings of extratemporal lobe epilepsy patients. Cross-frequency coupling between low and high frequency oscillations in conjunction with an unsupervised learning algorithm - namely, hidden Markov models - was used to objectively identify seizure and non-seizure states as well as transition states. Channels consistently capturing two and/or three distinct states in a 32-channel iEEG array were able to identify regions of interest located in resected tissue of patients who experienced improved post-surgical outcomes.

Keywords

electroencephalography; hidden Markov models; neurophysiology; unsupervised learning; cross-frequency coupling; extratemporal lobe epilepsy patients; hidden Markov models; human iEEG; intracranial electroencephalogram; modulated high frequency oscillations; nonseizure state transitions; seizure state; unsupervised learning algorithm; Computational modeling; Electrodes; Frequency modulation; Hidden Markov models; Oscillators; Training;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Engineering (NER), 2015 7th International IEEE/EMBS Conference on

Conference_Location

Montpellier

Type

conf

DOI

10.1109/NER.2015.7146777

Filename

7146777