Time-Frequency Detection of EEG Abnormalities

Abstract

The patterns obtained by a TF analysis of newborn EEG seizure signals show a linear FM or piecewise linear FM characteristic. This suggests a method of seizure detection and classification in the TF domain. A TF detector is proposed that involves cross-correlating the TFD of the EEG signal with a template. The design of the template takes into account the TF characteristics of the EEG seizure extracted in the TF domain. The performance of this time-frequency detector was tested on synthetic signals, corresponding to one specific type of seizure pattern (LFM). At the time of publication, the methodology was being extended to deal with LFM patterns of varying slopes, and with piecewise linear FM patterns. The procedure will then allow classification within the selected sub-classes. Another time-frequency approach to newborn EEG seizure detection is described in [11].