Detecting abnormal electroencephalograms using deep convolutional networks

K. G. van Leeuwen, H. Sun, M. Tabaeizadeh, A. F. Struck, M. J.A.M. van Putten, M. B. Westover

Research output: Contribution to journalArticleAcademicpeer-review

2 Citations (Scopus)

Abstract

Objectives: Electroencephalography (EEG) is a central part of the medical evaluation for patients with neurological disorders. Training an algorithm to label the EEG normal vs abnormal seems challenging, because of EEG heterogeneity and dependence of contextual factors, including age and sleep stage. Our objectives were to validate prior work on an independent data set suggesting that deep learning methods can discriminate between normal vs abnormal EEGs, to understand whether age and sleep stage information can improve discrimination, and to understand what factors lead to errors. Methods: We train a deep convolutional neural network on a heterogeneous set of 8522 routine EEGs from the Massachusetts General Hospital. We explore several strategies for optimizing model performance, including accounting for age and sleep stage. Results: The area under the receiver operating characteristic curve (AUC) on an independent test set (n = 851) is 0.917 marginally improved by including age (AUC = 0.924), and both age and sleep stages (AUC = 0.925), though not statistically significant. Conclusions: The model architecture generalizes well to an independent dataset. Adding age and sleep stage to the model does not significantly improve performance. Significance: Insights learned from misclassified examples, and minimal improvement by adding sleep stage and age suggest fruitful directions for further research.

Original languageEnglish
Pages (from-to)77-84
Number of pages8
JournalClinical neurophysiology
Volume130
Issue number1
Early online date17 Nov 2018
DOIs
Publication statusPublished - 1 Jan 2019

Fingerprint

Sleep Stages
Electroencephalography
Area Under Curve
Age Factors
Nervous System Diseases
ROC Curve
General Hospitals
Learning
Research

Keywords

  • Clinical neurophysiology
  • Computer aided diagnosis (CAD)
  • Convolutional neural networks (CNN)
  • Deep learning
  • Electroencephalograms (EEG)
  • Epilepsy

Cite this

van Leeuwen, K. G. ; Sun, H. ; Tabaeizadeh, M. ; Struck, A. F. ; van Putten, M. J.A.M. ; Westover, M. B. / Detecting abnormal electroencephalograms using deep convolutional networks. In: Clinical neurophysiology. 2019 ; Vol. 130, No. 1. pp. 77-84.
@article{9b802768535d4b159e3d365c3145576e,
title = "Detecting abnormal electroencephalograms using deep convolutional networks",
abstract = "Objectives: Electroencephalography (EEG) is a central part of the medical evaluation for patients with neurological disorders. Training an algorithm to label the EEG normal vs abnormal seems challenging, because of EEG heterogeneity and dependence of contextual factors, including age and sleep stage. Our objectives were to validate prior work on an independent data set suggesting that deep learning methods can discriminate between normal vs abnormal EEGs, to understand whether age and sleep stage information can improve discrimination, and to understand what factors lead to errors. Methods: We train a deep convolutional neural network on a heterogeneous set of 8522 routine EEGs from the Massachusetts General Hospital. We explore several strategies for optimizing model performance, including accounting for age and sleep stage. Results: The area under the receiver operating characteristic curve (AUC) on an independent test set (n = 851) is 0.917 marginally improved by including age (AUC = 0.924), and both age and sleep stages (AUC = 0.925), though not statistically significant. Conclusions: The model architecture generalizes well to an independent dataset. Adding age and sleep stage to the model does not significantly improve performance. Significance: Insights learned from misclassified examples, and minimal improvement by adding sleep stage and age suggest fruitful directions for further research.",
keywords = "Clinical neurophysiology, Computer aided diagnosis (CAD), Convolutional neural networks (CNN), Deep learning, Electroencephalograms (EEG), Epilepsy",
author = "{van Leeuwen}, {K. G.} and H. Sun and M. Tabaeizadeh and Struck, {A. F.} and {van Putten}, {M. J.A.M.} and Westover, {M. B.}",
year = "2019",
month = "1",
day = "1",
doi = "10.1016/j.clinph.2018.10.012",
language = "English",
volume = "130",
pages = "77--84",
journal = "Clinical neurophysiology",
issn = "1388-2457",
publisher = "Elsevier",
number = "1",

}

Detecting abnormal electroencephalograms using deep convolutional networks. / van Leeuwen, K. G.; Sun, H.; Tabaeizadeh, M.; Struck, A. F.; van Putten, M. J.A.M.; Westover, M. B.

In: Clinical neurophysiology, Vol. 130, No. 1, 01.01.2019, p. 77-84.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Detecting abnormal electroencephalograms using deep convolutional networks

AU - van Leeuwen, K. G.

AU - Sun, H.

AU - Tabaeizadeh, M.

AU - Struck, A. F.

AU - van Putten, M. J.A.M.

AU - Westover, M. B.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Objectives: Electroencephalography (EEG) is a central part of the medical evaluation for patients with neurological disorders. Training an algorithm to label the EEG normal vs abnormal seems challenging, because of EEG heterogeneity and dependence of contextual factors, including age and sleep stage. Our objectives were to validate prior work on an independent data set suggesting that deep learning methods can discriminate between normal vs abnormal EEGs, to understand whether age and sleep stage information can improve discrimination, and to understand what factors lead to errors. Methods: We train a deep convolutional neural network on a heterogeneous set of 8522 routine EEGs from the Massachusetts General Hospital. We explore several strategies for optimizing model performance, including accounting for age and sleep stage. Results: The area under the receiver operating characteristic curve (AUC) on an independent test set (n = 851) is 0.917 marginally improved by including age (AUC = 0.924), and both age and sleep stages (AUC = 0.925), though not statistically significant. Conclusions: The model architecture generalizes well to an independent dataset. Adding age and sleep stage to the model does not significantly improve performance. Significance: Insights learned from misclassified examples, and minimal improvement by adding sleep stage and age suggest fruitful directions for further research.

AB - Objectives: Electroencephalography (EEG) is a central part of the medical evaluation for patients with neurological disorders. Training an algorithm to label the EEG normal vs abnormal seems challenging, because of EEG heterogeneity and dependence of contextual factors, including age and sleep stage. Our objectives were to validate prior work on an independent data set suggesting that deep learning methods can discriminate between normal vs abnormal EEGs, to understand whether age and sleep stage information can improve discrimination, and to understand what factors lead to errors. Methods: We train a deep convolutional neural network on a heterogeneous set of 8522 routine EEGs from the Massachusetts General Hospital. We explore several strategies for optimizing model performance, including accounting for age and sleep stage. Results: The area under the receiver operating characteristic curve (AUC) on an independent test set (n = 851) is 0.917 marginally improved by including age (AUC = 0.924), and both age and sleep stages (AUC = 0.925), though not statistically significant. Conclusions: The model architecture generalizes well to an independent dataset. Adding age and sleep stage to the model does not significantly improve performance. Significance: Insights learned from misclassified examples, and minimal improvement by adding sleep stage and age suggest fruitful directions for further research.

KW - Clinical neurophysiology

KW - Computer aided diagnosis (CAD)

KW - Convolutional neural networks (CNN)

KW - Deep learning

KW - Electroencephalograms (EEG)

KW - Epilepsy

UR - http://www.scopus.com/inward/record.url?scp=85057139469&partnerID=8YFLogxK

U2 - 10.1016/j.clinph.2018.10.012

DO - 10.1016/j.clinph.2018.10.012

M3 - Article

VL - 130

SP - 77

EP - 84

JO - Clinical neurophysiology

JF - Clinical neurophysiology

SN - 1388-2457

IS - 1

ER -