Identifying signal-dependent information about the preictal state: A comparison across ECoG, EEG and EKG using deep learning.
Deep neural networks
Electrocardiogram
Epilepsy
Precision medicine
Seizure prediction
Journal
EBioMedicine
ISSN: 2352-3964
Titre abrégé: EBioMedicine
Pays: Netherlands
ID NLM: 101647039
Informations de publication
Date de publication:
Jul 2019
Jul 2019
Historique:
received:
30
04
2019
revised:
08
06
2019
accepted:
01
07
2019
pubmed:
14
7
2019
medline:
18
12
2019
entrez:
14
7
2019
Statut:
ppublish
Résumé
The inability to reliably assess seizure risk is a major burden for epilepsy patients and prevents developing better treatments. Recent advances have paved the way for increasingly accurate seizure preictal state detection algorithms, primarily using electrocorticography (ECoG). To develop seizure forecasting for broad clinical and ambulatory use, however, less complex and invasive modalities are needed. Algorithms using scalp electroencephalography (EEG) and electrocardiography (EKG) have also achieved better than chance performance. But it remains unknown how much preictal information is in ECoG versus modalities amenable to everyday use - such as EKG and single channel EEG - and how to optimally extract that preictal information for seizure prediction. We apply deep learning - a powerful method to extract information from complex data - on a large epilepsy data set containing multi-day, simultaneous recordings of EKG, ECoG, and EEG, using a variety of feature sets. We use the relative performance of our algorithms to compare the preictal information contained in each modality. We find that single-channel EKG contains a comparable amount of preictal information as scalp EEG with up to 21 channels and that preictal information is best extracted not with standard heart rate measures, but from the power spectral density. We report that preictal information is not preferentially contained in EEG or ECoG channels within the seizure onset zone. Collectively, these insights may help to devise future prospective, minimally invasive long-term epilepsy monitoring trials with single-channel EKG as a particularly promising modality.
Sections du résumé
BACKGROUND
BACKGROUND
The inability to reliably assess seizure risk is a major burden for epilepsy patients and prevents developing better treatments. Recent advances have paved the way for increasingly accurate seizure preictal state detection algorithms, primarily using electrocorticography (ECoG). To develop seizure forecasting for broad clinical and ambulatory use, however, less complex and invasive modalities are needed. Algorithms using scalp electroencephalography (EEG) and electrocardiography (EKG) have also achieved better than chance performance. But it remains unknown how much preictal information is in ECoG versus modalities amenable to everyday use - such as EKG and single channel EEG - and how to optimally extract that preictal information for seizure prediction.
METHODS
METHODS
We apply deep learning - a powerful method to extract information from complex data - on a large epilepsy data set containing multi-day, simultaneous recordings of EKG, ECoG, and EEG, using a variety of feature sets. We use the relative performance of our algorithms to compare the preictal information contained in each modality.
RESULTS
RESULTS
We find that single-channel EKG contains a comparable amount of preictal information as scalp EEG with up to 21 channels and that preictal information is best extracted not with standard heart rate measures, but from the power spectral density. We report that preictal information is not preferentially contained in EEG or ECoG channels within the seizure onset zone.
CONCLUSION
CONCLUSIONS
Collectively, these insights may help to devise future prospective, minimally invasive long-term epilepsy monitoring trials with single-channel EKG as a particularly promising modality.
Identifiants
pubmed: 31300348
pii: S2352-3964(19)30436-0
doi: 10.1016/j.ebiom.2019.07.001
pmc: PMC6642360
pii:
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
422-431Informations de copyright
Copyright © 2019. Published by Elsevier B.V.
Références
Med Biol Eng Comput. 2005 Mar;43(2):230-9
pubmed: 15865133
Brain. 2007 Feb;130(Pt 2):314-33
pubmed: 17008335
Brain. 2009 Apr;132(Pt 4):1013-21
pubmed: 19251759
Epilepsy Curr. 2009 Jul-Aug;9(4):91-5
pubmed: 19693322
Comput Methods Programs Biomed. 2012 Jun;106(3):127-38
pubmed: 20863589
Epilepsy Behav. 2011 Dec;22(4):685-91
pubmed: 21975299
Epilepsy Behav. 2011 Dec;22 Suppl 1:S119-26
pubmed: 22078512
PLoS One. 2012;7(2):e30371
pubmed: 22363431
Epilepsia. 2012 Aug;53(8):1314-21
pubmed: 22642646
Lancet Neurol. 2013 Jun;12(6):563-71
pubmed: 23642342
IEEE Trans Pattern Anal Mach Intell. 2013 Aug;35(8):1798-828
pubmed: 23787338
Ann Indian Acad Neurol. 2014 Mar;17(Suppl 1):S12-7
pubmed: 24791078
Neural Netw. 2015 Jan;61:85-117
pubmed: 25462637
Int J Neural Syst. 2015 Aug;25(5):1550018
pubmed: 25986753
PLoS One. 2015 Jul 10;10(7):e0130140
pubmed: 26161953
Europace. 2015 Sep;17(9):1341-53
pubmed: 26177817
Proc Natl Acad Sci U S A. 2015 Nov 24;112(47):14694-9
pubmed: 26554021
IEEE Trans Biomed Eng. 2016 Jun;63(6):1321-32
pubmed: 26841385
Brain. 2016 Apr;139(Pt 4):1066-78
pubmed: 26912639
Commun Integr Biol. 2016 Jan 05;9(1):e1128598
pubmed: 27065159
Brain. 2016 Jun;139(Pt 6):1625-7
pubmed: 27234060
Curr Opin Neurol. 2017 Apr;30(2):167-173
pubmed: 28118302
Front Physiol. 2017 Oct 05;8:765
pubmed: 29051738
EBioMedicine. 2018 Jan;27:103-111
pubmed: 29262989
eNeuro. 2017 Dec 27;4(6):
pubmed: 29291239
Neural Netw. 2018 Sep;105:104-111
pubmed: 29793128
Nat Rev Neurol. 2018 Oct;14(10):618-630
pubmed: 30131521
Eur Heart J. 1996 Mar;17(3):354-81
pubmed: 8737210