Forecasting Seizures Using Intracranial EEG Measures and SVM in Naturally Occurring Canine Epilepsy

Abstract

Management of drug resistant focal epilepsy would be greatly assisted by a reliable warning system capable of alerting patients prior to seizures to allow the patient to adjust activities or medication. Such a system requires successful identification of a preictal, or seizure-prone state. Identification of preictal states in continuous long- duration intracranial electroencephalographic (iEEG) recordings of dogs with naturally occurring epilepsy was investigated using a support vector machine (SVM) algorithm. The dogs studied were implanted with a 16-channel ambulatory iEEG recording device with average channel reference for a mean (st. dev.) of 380.4 (+87.5) days producing 220.2 (+104.1) days of intracranial EEG recorded at 400 Hz for analysis. The iEEG records had 51.6 (+52.8) seizures identified, of which 35.8 (+30.4) seizures were preceded by more than 4 hours of seizure-free data. Recorded iEEG data were stratified into 11 contiguous, non-overlapping frequency bands and binned into one-minute synchrony features for analysis. Performance of the SVM classifier was assessed using a 5-fold cross validation approach, where preictal training data were taken from 90 minute windows with a 5 minute pre-seizure offset. Analysis of the optimal preictal training time was performed by repeating the cross validation over a range of preictal windows and comparing results. We show that the optimization of feature selection varies for each subject, i.e. algorithms are subject specific, but achieve prediction performance significantly better than a time-matched Poisson random predictor (p<0.05) in 5/5 dogs analyzed.

Towards Improved Design and Evaluation of Epileptic Seizure Predictors

“…The contestants used a variety of machine learning models applied to a wide range of features from the time and frequency domains. The general approach, however, was much the same as in many previous studies, confirming that spectral power in discrete frequency bands is a valuable feature for seizure forecasting [5], [6], [11]. Moreover, SVM [12] was the most commonly used algorithm, which follows the trends in the seizure prediction research community [13].…”

“…In many works, seizure prediction models are trained to classify segments of a fixed length [6], [19]- [21]. In this case, once the preictal probability exceeds a certain threshold, the seizure forecasting system triggers an alarm, and the warning state persists for the same duration as the length of the classified segment.…”

“…In this case, once the preictal probability exceeds a certain threshold, the seizure forecasting system triggers an alarm, and the warning state persists for the same duration as the length of the classified segment. The seizure is considered as predicted if it occurs while the alarm is on [6]. From a machine learning perspective, this formulation makes the problem very well defined.…”

“…Recently, several datasets of long-term intracranial EEG (iEEG) recordings from canine subjects have become available and found their usage in a few studies [5], [6]. The reason behind exploiting canine data is that epilepsy in dogs and humans is proven to be highly similar [7] and it is easier to obtain relatively long iEEG recordings from canines.…”

Predicting Tunnel Squeezing Using Multiclass Support Vector Machines

“…SVMs do not require prior knowledge of a particular model form, possess a flexible nonlinear modeling capability, and have high generalization performance. erefore, they have become popular in various fields, such as mechanical engineering [26], biomedical engineering [27,28], information and communication engineering [29], and agriculture [30].…”