The performance of traditional linear (variance based) methods for the identification and prediction of epileptic seizures are contrasted with "modern" methods from nonlinear time series analysis. We note several flaws of design in demonstrations claiming to establish the efficacy of nonlinear techniques; in particular, we examine published evidence for precursor identification. We perform null hypothesis tests using relevant surrogate data to demonstrate that decreases in the correlation density prior to and during seizure may simply reflect increases in the variance.

Original publication

DOI

10.1109/TBME.2003.810688

Type

Journal article

Journal

IEEE Trans Biomed Eng

Publication Date

05/2003

Volume

50

Pages

628 - 633

Keywords

Algorithms, Brain, Computer Simulation, Electrodes, Implanted, Electroencephalography, Epilepsy, Temporal Lobe, False Positive Reactions, Humans, Linear Models, Models, Neurological, Nonlinear Dynamics, Quality Control, Reproducibility of Results, Sclerosis, Seizures, Sensitivity and Specificity, Signal Processing, Computer-Assisted, Statistics as Topic