Time-frequency extraction of EEG spike events for seizure detection in neonate

Abstract

There are a number of approaches for analysing EEG signals in the time, frequency, and time-frequency domains. However due to the nonstationarity of the EEG signals, the time-frequency methods proved to be superior. This paper presents a new method for the detection of newborn EEG seizure activity in the time-frequency domain. The proposed approach utilises 30-second epochs of EEG signal and analyses one frequency slice of their time-frequency representations at about 75 Hz. Spiking activity in these high frequency slices are used to distinguish between seizure and nonseizure activities. Using histograms of intervals between successive spikes, we were able to show the dramatic difference between spikes originating from seizure and those from nonseizure. This finding led to this novel method of seizure detection in newborn babies.