Seizure pattern-specific epileptic epoch detection in patients with intellectual disability

L. Wang, J.B.A.M. Arends, X. Long, P.J.M. Cluitmans, J.P. van Dijk

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

6 Citaties (Scopus)
5 Downloads (Pure)

Uittreksel

Electroencephalogram (EEG) features are crucial for the seizure detection performance. Traditional algorithms are designed for a population with normal brain development. However, for patients with an intellectual disability the seizure detection performance is still largely unknown. In addition, distinct EEG activities/patterns occur during the evolution of seizure events. However, few studies distinguished what EEG activities contribute to accurate seizure detections. To evaluate the effect of different seizure patterns on the seizure detection, we start from the four predefined seizure patterns: wave, fast spike, spike-wave complex, and seizure-related EMG artifacts. A wide range of promising EEG features in the time, frequency, time–frequency, and spatio-temporal domains, as well as synchronization-based features were extracted to characterize these patterns. The performance of seizure detection was evaluated in an epoch-based way. EEG recordings of 615 h from 29 epilepsy patients with intellectual disability were used in this study for validation. Results show that the seizure patterns of wave, and seizure-related EMG were easier to detect than the fast spike, spike-wave patterns, with sensitivities of 0.76, 0.74, 0.42, and 0.51, respectively (when specificity approximately equal to 1). We achieved the overall epoch-based detection performance with sensitivity of 68%, positive predictive value (PPV) 81%, and average duration of false detection 0.76 s per hour. Feature importance analysis indicated that the classification performance of traditional EEG features can be improved when combined with our newly-proposed features from the spatio-temporal domain and the synchronization-based methods.
Originele taal-2Engels
Pagina's (van-tot)38-49
Aantal pagina's12
TijdschriftBiomedical Signal Processing and Control
Volume35
DOI's
StatusGepubliceerd - mrt 2017

Vingerafdruk

Electroencephalography
Intellectual Disability
Seizures
Synchronization
Brain
Validation Studies
Artifacts
Epilepsy

Citeer dit

@article{f727e6a6dd89451792710474b7f47900,
title = "Seizure pattern-specific epileptic epoch detection in patients with intellectual disability",
abstract = "Electroencephalogram (EEG) features are crucial for the seizure detection performance. Traditional algorithms are designed for a population with normal brain development. However, for patients with an intellectual disability the seizure detection performance is still largely unknown. In addition, distinct EEG activities/patterns occur during the evolution of seizure events. However, few studies distinguished what EEG activities contribute to accurate seizure detections. To evaluate the effect of different seizure patterns on the seizure detection, we start from the four predefined seizure patterns: wave, fast spike, spike-wave complex, and seizure-related EMG artifacts. A wide range of promising EEG features in the time, frequency, time–frequency, and spatio-temporal domains, as well as synchronization-based features were extracted to characterize these patterns. The performance of seizure detection was evaluated in an epoch-based way. EEG recordings of 615 h from 29 epilepsy patients with intellectual disability were used in this study for validation. Results show that the seizure patterns of wave, and seizure-related EMG were easier to detect than the fast spike, spike-wave patterns, with sensitivities of 0.76, 0.74, 0.42, and 0.51, respectively (when specificity approximately equal to 1). We achieved the overall epoch-based detection performance with sensitivity of 68{\%}, positive predictive value (PPV) 81{\%}, and average duration of false detection 0.76 s per hour. Feature importance analysis indicated that the classification performance of traditional EEG features can be improved when combined with our newly-proposed features from the spatio-temporal domain and the synchronization-based methods.",
keywords = "EEG, Seizure detection, Seizure patterns, Classifiers, Feature importance, Intellectual disability",
author = "L. Wang and J.B.A.M. Arends and X. Long and P.J.M. Cluitmans and {van Dijk}, J.P.",
year = "2017",
month = "3",
doi = "10.1016/j.bspc.2017.02.008",
language = "English",
volume = "35",
pages = "38--49",
journal = "Biomedical Signal Processing and Control",
issn = "1746-8094",
publisher = "Elsevier",

}

Seizure pattern-specific epileptic epoch detection in patients with intellectual disability. / Wang, L.; Arends, J.B.A.M.; Long, X.; Cluitmans, P.J.M.; van Dijk, J.P.

In: Biomedical Signal Processing and Control, Vol. 35, 03.2017, blz. 38-49.

Onderzoeksoutput: Bijdrage aan tijdschriftTijdschriftartikelAcademicpeer review

TY - JOUR

T1 - Seizure pattern-specific epileptic epoch detection in patients with intellectual disability

AU - Wang, L.

AU - Arends, J.B.A.M.

AU - Long, X.

AU - Cluitmans, P.J.M.

AU - van Dijk, J.P.

PY - 2017/3

Y1 - 2017/3

N2 - Electroencephalogram (EEG) features are crucial for the seizure detection performance. Traditional algorithms are designed for a population with normal brain development. However, for patients with an intellectual disability the seizure detection performance is still largely unknown. In addition, distinct EEG activities/patterns occur during the evolution of seizure events. However, few studies distinguished what EEG activities contribute to accurate seizure detections. To evaluate the effect of different seizure patterns on the seizure detection, we start from the four predefined seizure patterns: wave, fast spike, spike-wave complex, and seizure-related EMG artifacts. A wide range of promising EEG features in the time, frequency, time–frequency, and spatio-temporal domains, as well as synchronization-based features were extracted to characterize these patterns. The performance of seizure detection was evaluated in an epoch-based way. EEG recordings of 615 h from 29 epilepsy patients with intellectual disability were used in this study for validation. Results show that the seizure patterns of wave, and seizure-related EMG were easier to detect than the fast spike, spike-wave patterns, with sensitivities of 0.76, 0.74, 0.42, and 0.51, respectively (when specificity approximately equal to 1). We achieved the overall epoch-based detection performance with sensitivity of 68%, positive predictive value (PPV) 81%, and average duration of false detection 0.76 s per hour. Feature importance analysis indicated that the classification performance of traditional EEG features can be improved when combined with our newly-proposed features from the spatio-temporal domain and the synchronization-based methods.

AB - Electroencephalogram (EEG) features are crucial for the seizure detection performance. Traditional algorithms are designed for a population with normal brain development. However, for patients with an intellectual disability the seizure detection performance is still largely unknown. In addition, distinct EEG activities/patterns occur during the evolution of seizure events. However, few studies distinguished what EEG activities contribute to accurate seizure detections. To evaluate the effect of different seizure patterns on the seizure detection, we start from the four predefined seizure patterns: wave, fast spike, spike-wave complex, and seizure-related EMG artifacts. A wide range of promising EEG features in the time, frequency, time–frequency, and spatio-temporal domains, as well as synchronization-based features were extracted to characterize these patterns. The performance of seizure detection was evaluated in an epoch-based way. EEG recordings of 615 h from 29 epilepsy patients with intellectual disability were used in this study for validation. Results show that the seizure patterns of wave, and seizure-related EMG were easier to detect than the fast spike, spike-wave patterns, with sensitivities of 0.76, 0.74, 0.42, and 0.51, respectively (when specificity approximately equal to 1). We achieved the overall epoch-based detection performance with sensitivity of 68%, positive predictive value (PPV) 81%, and average duration of false detection 0.76 s per hour. Feature importance analysis indicated that the classification performance of traditional EEG features can be improved when combined with our newly-proposed features from the spatio-temporal domain and the synchronization-based methods.

KW - EEG

KW - Seizure detection

KW - Seizure patterns

KW - Classifiers

KW - Feature importance

KW - Intellectual disability

U2 - 10.1016/j.bspc.2017.02.008

DO - 10.1016/j.bspc.2017.02.008

M3 - Article

VL - 35

SP - 38

EP - 49

JO - Biomedical Signal Processing and Control

JF - Biomedical Signal Processing and Control

SN - 1746-8094

ER -