Automatic detection of generalized epilepsy through multivariate singular spectral analysis

Abstract

The collection and interpretation of electroencephalogram (EEG) signals are laborious and time-consuming activities, requiring a trained specialist to perform them. Automatic detection of epilepsy may be a solution. However, research on the subject has focused on detecting specific, non-generalized epilepsies in a larger patient population. Decomposition of signals, through singular spectrum analysis, of records of patients with epilepsy for subsequent verification of the energy limit. These records were available in a publicly accessible signal bank. The use of different weights to calculate means and standard deviations of the energy series and different sample sizes contributed to improve the diagnosis.