Abstract

It is often difficult to predict seizure recurrence in subjects who have suffered a first-ever epileptic seizure. In this study, the predictive value of physiological signals measured using Electroencephalography (EEG) and functional MRI (fMRI) is assessed. In particular those patients developing epilepsy (i.e. a second unprovoked seizure) that were initially evaluated as having a low risk of seizure recurrence are of interest.In total, 26 epilepsy patients, of which 8 were initially evaluated as having a low risk of seizure recurrence (i.e. converters), and 17 subjects with only a single seizure were included. All subjects underwent routine EEG as well as fMRI measurements. For diagnostic classification, features related to the temporal dynamics were determined for both the processed EEG and fMRI data. Subsequently, a logistic regression classifier was trained on epilepsy and first-seizure subjects. The trained model was tested using the clinically relevant converters group.The sensitivity, specificity, and AUC (mean +/- SD) of the regression model including metrics from both modalities were 74 +/- 19%, 82 +/- 18%, and 0.75 +/- 0.12, respectively. Positive and negative predictive values (mean SD) of the regression model with both EEG and fMRI features are 84 +/- 14% and 78 +/- 12%. Moreover, this EEG/fMRI model showed significant improvements compared to the clinical diagnosis, whereas the models using metrics from either EEG or fMRI do not reach significance (p > 0.05).Temporal metrics computationally derived from EEG and fMRI time signals may clinically aid and synergistically improve the predictive value in a first-seizure sample. (C) 2020 The Author(s). Published by Elsevier Inc.
Original languageEnglish
Article number107651
Number of pages8
JournalEpilepsy & Behavior
Volume115
DOIs
Publication statusPublished - 1 Feb 2021

Keywords

  • Electroencephalography
  • First fit
  • New-onset epilepsy
  • fMRI
  • RESTING-STATE FMRI
  • CONNECTIVITY
  • ELECTROENCEPHALOGRAPHY
  • DYNAMICS
  • ADULTS
  • NETWORKS
  • CHILDREN

Cite this