Semi-supervised training data selection improves seizure forecasting in canines with epilepsy

Mona Nasseri, Vaclav Kremen, Petr Nejedly, Inyong Kim, Su Youne Chang, Hang Joon Jo, Hari Guragain, Nathaniel Nelson, Edward Patterson, Beverly K. Sturges, Chelsea M. Crowe, Tim Denison, Benjamin H. Brinkmann, Gregory A. Worrell

Research output: Contribution to journalArticle

Abstract

Objective: Conventional selection of pre-ictal EEG epochs for seizure prediction algorithm training data typically assumes a continuous pre-ictal brain state preceding a seizure. This is carried out by defining a fixed duration, pre-ictal time period before seizures from which pre-ictal training data epochs are uniformly sampled. However, stochastic physiological and pathological fluctuations in EEG data characteristics and underlying brain states suggest that pre-ictal state dynamics may be more complex, and selection of pre-ictal training data segments to reflect this could improve algorithm performance. Methods: We propose a semi-supervised technique to select pre-ictal training data most distinguishable from interictal EEG according to pre-specified data characteristics. The proposed method uses hierarchical clustering to identify optimal pre-ictal data epochs. Results: In this paper we compare the performance of a seizure forecasting algorithm with and without hierarchical clustering of pre-ictal periods in chronic iEEG recordings from six canines with naturally occurring epilepsy. Hierarchical clustering of training data improved results for Time In Warning (TIW) (0.18 vs. 0.23) and False Positive Rate (FPR) (0.5 vs. 0.59) when evaluated across all subjects (p < 0.001, n = 6). Results were mixed when evaluating TIW, FPR, and Sensitivity for individual dogs. Conclusion: Hierarchical clustering is a helpful method for training data selection overall, but should be evaluated on a subject-wise basis. Significance: The clustering method can be used to optimize results of forecasting towards sensitivity or TIW or FPR, and therefore can be useful for epilepsy management.

Original languageEnglish (US)
Article number101743
JournalBiomedical Signal Processing and Control
Volume57
DOIs
StatePublished - Mar 2020

Fingerprint

Electroencephalography
Canidae
Epilepsy
Seizures
Stroke
Brain
Cluster Analysis
Dogs

Keywords

  • Hierarchical clustering
  • Machine learning
  • Seizure forecasting

Cite this

Nasseri, M., Kremen, V., Nejedly, P., Kim, I., Chang, S. Y., Jo, H. J., ... Worrell, G. A. (2020). Semi-supervised training data selection improves seizure forecasting in canines with epilepsy. Biomedical Signal Processing and Control, 57, [101743]. https://doi.org/10.1016/j.bspc.2019.101743

Semi-supervised training data selection improves seizure forecasting in canines with epilepsy. / Nasseri, Mona; Kremen, Vaclav; Nejedly, Petr; Kim, Inyong; Chang, Su Youne; Jo, Hang Joon; Guragain, Hari; Nelson, Nathaniel; Patterson, Edward; Sturges, Beverly K.; Crowe, Chelsea M.; Denison, Tim; Brinkmann, Benjamin H.; Worrell, Gregory A.

In: Biomedical Signal Processing and Control, Vol. 57, 101743, 03.2020.

Research output: Contribution to journalArticle

Nasseri, M, Kremen, V, Nejedly, P, Kim, I, Chang, SY, Jo, HJ, Guragain, H, Nelson, N, Patterson, E, Sturges, BK, Crowe, CM, Denison, T, Brinkmann, BH & Worrell, GA 2020, 'Semi-supervised training data selection improves seizure forecasting in canines with epilepsy', Biomedical Signal Processing and Control, vol. 57, 101743. https://doi.org/10.1016/j.bspc.2019.101743
Nasseri, Mona ; Kremen, Vaclav ; Nejedly, Petr ; Kim, Inyong ; Chang, Su Youne ; Jo, Hang Joon ; Guragain, Hari ; Nelson, Nathaniel ; Patterson, Edward ; Sturges, Beverly K. ; Crowe, Chelsea M. ; Denison, Tim ; Brinkmann, Benjamin H. ; Worrell, Gregory A. / Semi-supervised training data selection improves seizure forecasting in canines with epilepsy. In: Biomedical Signal Processing and Control. 2020 ; Vol. 57.
@article{99fd7e4cb02d4e7da25f8bfb3bcbeb62,
title = "Semi-supervised training data selection improves seizure forecasting in canines with epilepsy",
abstract = "Objective: Conventional selection of pre-ictal EEG epochs for seizure prediction algorithm training data typically assumes a continuous pre-ictal brain state preceding a seizure. This is carried out by defining a fixed duration, pre-ictal time period before seizures from which pre-ictal training data epochs are uniformly sampled. However, stochastic physiological and pathological fluctuations in EEG data characteristics and underlying brain states suggest that pre-ictal state dynamics may be more complex, and selection of pre-ictal training data segments to reflect this could improve algorithm performance. Methods: We propose a semi-supervised technique to select pre-ictal training data most distinguishable from interictal EEG according to pre-specified data characteristics. The proposed method uses hierarchical clustering to identify optimal pre-ictal data epochs. Results: In this paper we compare the performance of a seizure forecasting algorithm with and without hierarchical clustering of pre-ictal periods in chronic iEEG recordings from six canines with naturally occurring epilepsy. Hierarchical clustering of training data improved results for Time In Warning (TIW) (0.18 vs. 0.23) and False Positive Rate (FPR) (0.5 vs. 0.59) when evaluated across all subjects (p < 0.001, n = 6). Results were mixed when evaluating TIW, FPR, and Sensitivity for individual dogs. Conclusion: Hierarchical clustering is a helpful method for training data selection overall, but should be evaluated on a subject-wise basis. Significance: The clustering method can be used to optimize results of forecasting towards sensitivity or TIW or FPR, and therefore can be useful for epilepsy management.",
keywords = "Hierarchical clustering, Machine learning, Seizure forecasting",
author = "Mona Nasseri and Vaclav Kremen and Petr Nejedly and Inyong Kim and Chang, {Su Youne} and Jo, {Hang Joon} and Hari Guragain and Nathaniel Nelson and Edward Patterson and Sturges, {Beverly K.} and Crowe, {Chelsea M.} and Tim Denison and Brinkmann, {Benjamin H.} and Worrell, {Gregory A.}",
year = "2020",
month = "3",
doi = "10.1016/j.bspc.2019.101743",
language = "English (US)",
volume = "57",
journal = "Biomedical Signal Processing and Control",
issn = "1746-8094",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Semi-supervised training data selection improves seizure forecasting in canines with epilepsy

AU - Nasseri, Mona

AU - Kremen, Vaclav

AU - Nejedly, Petr

AU - Kim, Inyong

AU - Chang, Su Youne

AU - Jo, Hang Joon

AU - Guragain, Hari

AU - Nelson, Nathaniel

AU - Patterson, Edward

AU - Sturges, Beverly K.

AU - Crowe, Chelsea M.

AU - Denison, Tim

AU - Brinkmann, Benjamin H.

AU - Worrell, Gregory A.

PY - 2020/3

Y1 - 2020/3

N2 - Objective: Conventional selection of pre-ictal EEG epochs for seizure prediction algorithm training data typically assumes a continuous pre-ictal brain state preceding a seizure. This is carried out by defining a fixed duration, pre-ictal time period before seizures from which pre-ictal training data epochs are uniformly sampled. However, stochastic physiological and pathological fluctuations in EEG data characteristics and underlying brain states suggest that pre-ictal state dynamics may be more complex, and selection of pre-ictal training data segments to reflect this could improve algorithm performance. Methods: We propose a semi-supervised technique to select pre-ictal training data most distinguishable from interictal EEG according to pre-specified data characteristics. The proposed method uses hierarchical clustering to identify optimal pre-ictal data epochs. Results: In this paper we compare the performance of a seizure forecasting algorithm with and without hierarchical clustering of pre-ictal periods in chronic iEEG recordings from six canines with naturally occurring epilepsy. Hierarchical clustering of training data improved results for Time In Warning (TIW) (0.18 vs. 0.23) and False Positive Rate (FPR) (0.5 vs. 0.59) when evaluated across all subjects (p < 0.001, n = 6). Results were mixed when evaluating TIW, FPR, and Sensitivity for individual dogs. Conclusion: Hierarchical clustering is a helpful method for training data selection overall, but should be evaluated on a subject-wise basis. Significance: The clustering method can be used to optimize results of forecasting towards sensitivity or TIW or FPR, and therefore can be useful for epilepsy management.

AB - Objective: Conventional selection of pre-ictal EEG epochs for seizure prediction algorithm training data typically assumes a continuous pre-ictal brain state preceding a seizure. This is carried out by defining a fixed duration, pre-ictal time period before seizures from which pre-ictal training data epochs are uniformly sampled. However, stochastic physiological and pathological fluctuations in EEG data characteristics and underlying brain states suggest that pre-ictal state dynamics may be more complex, and selection of pre-ictal training data segments to reflect this could improve algorithm performance. Methods: We propose a semi-supervised technique to select pre-ictal training data most distinguishable from interictal EEG according to pre-specified data characteristics. The proposed method uses hierarchical clustering to identify optimal pre-ictal data epochs. Results: In this paper we compare the performance of a seizure forecasting algorithm with and without hierarchical clustering of pre-ictal periods in chronic iEEG recordings from six canines with naturally occurring epilepsy. Hierarchical clustering of training data improved results for Time In Warning (TIW) (0.18 vs. 0.23) and False Positive Rate (FPR) (0.5 vs. 0.59) when evaluated across all subjects (p < 0.001, n = 6). Results were mixed when evaluating TIW, FPR, and Sensitivity for individual dogs. Conclusion: Hierarchical clustering is a helpful method for training data selection overall, but should be evaluated on a subject-wise basis. Significance: The clustering method can be used to optimize results of forecasting towards sensitivity or TIW or FPR, and therefore can be useful for epilepsy management.

KW - Hierarchical clustering

KW - Machine learning

KW - Seizure forecasting

UR - http://www.scopus.com/inward/record.url?scp=85074854563&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85074854563&partnerID=8YFLogxK

U2 - 10.1016/j.bspc.2019.101743

DO - 10.1016/j.bspc.2019.101743

M3 - Article

AN - SCOPUS:85074854563

VL - 57

JO - Biomedical Signal Processing and Control

JF - Biomedical Signal Processing and Control

SN - 1746-8094

M1 - 101743

ER -