Seizure detection of newborn EEG using a model based approach: a review of performance

This investigation looks into previous work performed by Roessgen et al. (1998) on the modelling of neonatal electroencephalogram (EEG) seizure signals and the utilization of parameter estimation as a means of seizure detection. The aim is to make precise a number of important results and to provide further details on several aspects of the previous publication. Upon reimplementation of the proposed method for experimental purposes a number of errors in the previously published equations were found. In particular it was found that the Power Spectral Density (PSD) equation for the Seizure EEG Model (SEM) proposed by Roessgen et al. was published incorrectly. A number of errors were also found in the derivation of the PSD for the random seizure input waveform. Besides this, the parametrisation of the SEM is redefined to ensure mathematical correctness. It is hoped that these improvements and appropriate additional comments will enhance the validity and reproducibility of the method.