A multi-feature and multi-channel univariate selection process for seizure prediction

Maryann D'Alessandro, George Vachtsevanos, Rosana Esteller, Javier Echauz, Stephen Cranstoun, Gregory Alan Worrell, Landi Parish, Brian Litt

Research output: Contribution to journalArticle

62 Citations (Scopus)

Abstract

Objective: To develop a prospective method for optimizing seizure prediction, given an array of implanted electrodes and a set of candidate quantitative features computed at each contact location. Methods: The method employs a genetic-based selection process, and then tunes a probabilistic neural network classifier to predict seizures within a 10 min prediction horizon. Initial seizure and interictal data were used for training, and the remaining IEEG data were used for testing. The method continues to train and learn over time. Results: Validation of these results over two workshop patients demonstrated a sensitivity of 100%, and 1.1 false positives per hour for Patient E, using a 2.4 s block predictor, and a failure of the method on Patient B. Conclusions: This study demonstrates a prospective, exploratory implementation of a seizure prediction method designed to adapt to individual patients with a wide variety of pre-ictal patterns, implanted electrodes and seizure types. Its current performance is limited likely by the small number of input channels and quantitative features employed in this study, and segmentation of the data set into training and testing sets rather than using all continuous data available. Significance: This technique theoretically has the potential to address the challenge presented by the heterogeneity of EEG patterns seen in medication-resistant epilepsy. A more comprehensive implementation utilizing all electrode sites, a broader feature library, and automated multi-feature fusion will be required to fully judge the method's potential for predicting seizures.

Original languageEnglish (US)
Pages (from-to)506-516
Number of pages11
JournalClinical Neurophysiology
Volume116
Issue number3
DOIs
StatePublished - Mar 2005

Fingerprint

Seizures
Implanted Electrodes
Libraries
Electroencephalography
Electrodes
Stroke
Education

Keywords

  • Classification
  • Feature extraction
  • Multiple channels
  • Multiple features
  • Seizure prediction

ASJC Scopus subject areas

  • Clinical Neurology
  • Radiology Nuclear Medicine and imaging
  • Neurology
  • Sensory Systems
  • Physiology (medical)

Cite this

D'Alessandro, M., Vachtsevanos, G., Esteller, R., Echauz, J., Cranstoun, S., Worrell, G. A., ... Litt, B. (2005). A multi-feature and multi-channel univariate selection process for seizure prediction. Clinical Neurophysiology, 116(3), 506-516. https://doi.org/10.1016/j.clinph.2004.11.014

A multi-feature and multi-channel univariate selection process for seizure prediction. / D'Alessandro, Maryann; Vachtsevanos, George; Esteller, Rosana; Echauz, Javier; Cranstoun, Stephen; Worrell, Gregory Alan; Parish, Landi; Litt, Brian.

In: Clinical Neurophysiology, Vol. 116, No. 3, 03.2005, p. 506-516.

Research output: Contribution to journalArticle

D'Alessandro, M, Vachtsevanos, G, Esteller, R, Echauz, J, Cranstoun, S, Worrell, GA, Parish, L & Litt, B 2005, 'A multi-feature and multi-channel univariate selection process for seizure prediction', Clinical Neurophysiology, vol. 116, no. 3, pp. 506-516. https://doi.org/10.1016/j.clinph.2004.11.014
D'Alessandro, Maryann ; Vachtsevanos, George ; Esteller, Rosana ; Echauz, Javier ; Cranstoun, Stephen ; Worrell, Gregory Alan ; Parish, Landi ; Litt, Brian. / A multi-feature and multi-channel univariate selection process for seizure prediction. In: Clinical Neurophysiology. 2005 ; Vol. 116, No. 3. pp. 506-516.
@article{0944f4fbef4d4b2a8e7319f40f49c010,
title = "A multi-feature and multi-channel univariate selection process for seizure prediction",
abstract = "Objective: To develop a prospective method for optimizing seizure prediction, given an array of implanted electrodes and a set of candidate quantitative features computed at each contact location. Methods: The method employs a genetic-based selection process, and then tunes a probabilistic neural network classifier to predict seizures within a 10 min prediction horizon. Initial seizure and interictal data were used for training, and the remaining IEEG data were used for testing. The method continues to train and learn over time. Results: Validation of these results over two workshop patients demonstrated a sensitivity of 100{\%}, and 1.1 false positives per hour for Patient E, using a 2.4 s block predictor, and a failure of the method on Patient B. Conclusions: This study demonstrates a prospective, exploratory implementation of a seizure prediction method designed to adapt to individual patients with a wide variety of pre-ictal patterns, implanted electrodes and seizure types. Its current performance is limited likely by the small number of input channels and quantitative features employed in this study, and segmentation of the data set into training and testing sets rather than using all continuous data available. Significance: This technique theoretically has the potential to address the challenge presented by the heterogeneity of EEG patterns seen in medication-resistant epilepsy. A more comprehensive implementation utilizing all electrode sites, a broader feature library, and automated multi-feature fusion will be required to fully judge the method's potential for predicting seizures.",
keywords = "Classification, Feature extraction, Multiple channels, Multiple features, Seizure prediction",
author = "Maryann D'Alessandro and George Vachtsevanos and Rosana Esteller and Javier Echauz and Stephen Cranstoun and Worrell, {Gregory Alan} and Landi Parish and Brian Litt",
year = "2005",
month = "3",
doi = "10.1016/j.clinph.2004.11.014",
language = "English (US)",
volume = "116",
pages = "506--516",
journal = "Clinical Neurophysiology",
issn = "1388-2457",
publisher = "Elsevier Ireland Ltd",
number = "3",

}

TY - JOUR

T1 - A multi-feature and multi-channel univariate selection process for seizure prediction

AU - D'Alessandro, Maryann

AU - Vachtsevanos, George

AU - Esteller, Rosana

AU - Echauz, Javier

AU - Cranstoun, Stephen

AU - Worrell, Gregory Alan

AU - Parish, Landi

AU - Litt, Brian

PY - 2005/3

Y1 - 2005/3

N2 - Objective: To develop a prospective method for optimizing seizure prediction, given an array of implanted electrodes and a set of candidate quantitative features computed at each contact location. Methods: The method employs a genetic-based selection process, and then tunes a probabilistic neural network classifier to predict seizures within a 10 min prediction horizon. Initial seizure and interictal data were used for training, and the remaining IEEG data were used for testing. The method continues to train and learn over time. Results: Validation of these results over two workshop patients demonstrated a sensitivity of 100%, and 1.1 false positives per hour for Patient E, using a 2.4 s block predictor, and a failure of the method on Patient B. Conclusions: This study demonstrates a prospective, exploratory implementation of a seizure prediction method designed to adapt to individual patients with a wide variety of pre-ictal patterns, implanted electrodes and seizure types. Its current performance is limited likely by the small number of input channels and quantitative features employed in this study, and segmentation of the data set into training and testing sets rather than using all continuous data available. Significance: This technique theoretically has the potential to address the challenge presented by the heterogeneity of EEG patterns seen in medication-resistant epilepsy. A more comprehensive implementation utilizing all electrode sites, a broader feature library, and automated multi-feature fusion will be required to fully judge the method's potential for predicting seizures.

AB - Objective: To develop a prospective method for optimizing seizure prediction, given an array of implanted electrodes and a set of candidate quantitative features computed at each contact location. Methods: The method employs a genetic-based selection process, and then tunes a probabilistic neural network classifier to predict seizures within a 10 min prediction horizon. Initial seizure and interictal data were used for training, and the remaining IEEG data were used for testing. The method continues to train and learn over time. Results: Validation of these results over two workshop patients demonstrated a sensitivity of 100%, and 1.1 false positives per hour for Patient E, using a 2.4 s block predictor, and a failure of the method on Patient B. Conclusions: This study demonstrates a prospective, exploratory implementation of a seizure prediction method designed to adapt to individual patients with a wide variety of pre-ictal patterns, implanted electrodes and seizure types. Its current performance is limited likely by the small number of input channels and quantitative features employed in this study, and segmentation of the data set into training and testing sets rather than using all continuous data available. Significance: This technique theoretically has the potential to address the challenge presented by the heterogeneity of EEG patterns seen in medication-resistant epilepsy. A more comprehensive implementation utilizing all electrode sites, a broader feature library, and automated multi-feature fusion will be required to fully judge the method's potential for predicting seizures.

KW - Classification

KW - Feature extraction

KW - Multiple channels

KW - Multiple features

KW - Seizure prediction

UR - http://www.scopus.com/inward/record.url?scp=13844298785&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=13844298785&partnerID=8YFLogxK

U2 - 10.1016/j.clinph.2004.11.014

DO - 10.1016/j.clinph.2004.11.014

M3 - Article

VL - 116

SP - 506

EP - 516

JO - Clinical Neurophysiology

JF - Clinical Neurophysiology

SN - 1388-2457

IS - 3

ER -