A comparative study of different nonlinear methods for localising seizure onset zone in epilepsy patients from intracranial electroencephalographic recordings.

Galán Augé, Cristian

A comparative study of different nonlinear methods for localising seizure onset zone in epilepsy patients from intracranial electroencephalographic recordings.

License

Citation

Galán Augé, C. A comparative study of different nonlinear methods for localising seizure onset zone in epilepsy patients from intracranial electroencephalographic recordings.. 2019. handle: http://hdl.handle.net/10230/44591

This citation was generated automatically.

Abstract

Most of the epilepsy cases are controlled using medication. Unfortunately, medication sometimes does not provide a sufficient control of seizures. For these cases, epilepsy surgery can be an option which needs the detection of the zone where seizures are generated. This involves measuring the electrical activity with electroencephalography (EEG). Intracranial EEG (iEEG) recordings allows a long-term monitoring and it can provide detailed information about the seizure onset zone (SOZ) in epilepsy patients. However, not all characteristics of the EEG can be captured by visual inspection, it may be time demanding and it has user variability. Thus, some analytic methods are used. In particular, in this study we apply and compare four measures in three different iEEG databases from epilepsy patients. These databases contains EEG recordings in the hemisphere where SOZ was located and in the contralateral hemisphere. Two of these methods test determinism in point processes by evaluating predictability: nonlinear prediction error (E) and rank-based nonlinear predictability score (S). The other two are symbolic methods: they analyse qualitative information rather than quantitative. These symbolic methods measure the occurrences of ordinal patterns, which are based on the ranking of a group of samples which are sorted by their amplitude. We found that the use of the predictability measures offers a better differentiation between focal and nonfocal signals than using the symbolic methods.