Feature Selection Techniques for the Analysis of Discriminative Features in Temporal and Frontal Lobe Epilepsy: A Comparative Study
Behrooz Abbaszadeh1, *, Cesar Alexandre Domingues Teixeira2, Mustapha C.E. Yagoub1
Identifiers and Pagination:Year: 2021
First Page: 1
Last Page: 15
Publisher Id: TOBEJ-15-1
Article History:Received Date: 15/12/2020
Revision Received Date: 25/3/2021
Acceptance Date: 14/4/2021
Electronic publication date: 07/06/2021
Collection year: 2021
open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: https://creativecommons.org/licenses/by/4.0/legalcode. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Because about 30% of epileptic patients suffer from refractory epilepsy, an efficient automatic seizure prediction tool is in great demand to improve their life quality.
In this work, time-domain discriminating preictal and interictal features were efficiently extracted from the intracranial electroencephalogram of twelve patients, i.e., six with temporal and six with frontal lobe epilepsy. The performance of three types of feature selection methods was compared using Matthews’s correlation coefficient (MCC).
Kruskal Wallis, a non-parametric approach, was found to perform better than the other approaches due to a simple and less resource consuming strategy as well as maintaining the highest MCC score. The impact of dividing the electroencephalogram signals into various sub-bands was investigated as well. The highest performance of Kruskal Wallis may suggest considering the importance of univariate features like complexity and interquartile ratio (IQR), along with autoregressive (AR) model parameters and the maximum (MAX) cross-correlation to efficiently predict epileptic seizures.
The proposed approach has the potential to be implemented on a low power device by considering a few simple time domain characteristics for a specific sub-band. It should be noted that, as there is not a great deal of literature on frontal lobe epilepsy, the results of this work can be considered promising.