Neuropediatrics 2006; 37 - P55
DOI: 10.1055/s-2006-974066

Effects of signal-to-noise ratio and averaging in non-invasive source localization of interictal EEG spikes

T Boppel 1, A Rupp 2, I Harting 3, K Hoechstetter 4, S Fauser 5, A Schulze-Bonhage 5, D Rating 1, M Scherg 4, T Bast 1
  • 1Universitätsklinik f. Kinder- u. Jugendmedizin, Pädiatrische Neurologie, Heidelberg, Germany
  • 2Universitätsklinik, Neurologie, Heidelberg, Germany
  • 3Universitätsklinik, Neuroradiologische Abteilung, Heidelberg, Germany
  • 4MEGIS Software GmbH, Gräfelfing, Germany
  • 5Universitätsklinik, Epilepsiezentrum, Freiburg, Germany

Aim: Source analysis estimates equivalent centres of an irritative zone, i.e. spiking grey matter. The cortical extent is difficult to assess from scalp EEG and remains unclear. Scatter of current dipoles in single spike analysis is frequently used as an indicator for extent. This study aims to reveal the influence of signal-to-noise ratio (SNR) on single spike analysis. Averaging a high number of „identical“ spikes increases SNR and allows onset analysis. However, only information on the maximum centre of all spike activity is provided since minimal differences get lost. We introduce a new method based on multiple subaveraging of „similar“ spikes (re-averaging) based on a bootstrapping.

Patients and methods: 25 patients with focal epilepsies (12m, 13 f, age 3–22yrs, mean 10yrs) were investigated by 32-channel EEG. A minimum of 100 highly correlating spikes (r>0.85, automatic pattern search) was detectable for 27 independent foci.100 spikes were randomly selected out of the total number of detected spikes. Dipole analysis was applied for defined time points (onset, peak) for each single spike and the average of all segments (BESA). A bootstrapping (repeated drawing and replacement) was applied for generating subaverages of 10, 25, 50 and 100 spike segments (each 1000 averages). Dipole analysis was performed for the given timepoints.

Results: Single dipole scatter was highly correlated with SNR (r=–0.83, p<0.0001) and was greatly reduced when analysing multiple subaverages of 10, 25, 50 and 100 spikes. In several subjects with known epileptogenic lesions the single spike scatter considerably exceeded the lesion and even false lateralisation was found in single cases. Onset analysis and differentiation of propagation phenomena were possible after multiple subaveraging.

Conclusion: SNR markedly influences the estimation of the extent of an activated cortical area. Scatter plots based on the dipole localization of randomly sampled subaverages improves dipole analysis by preserving information from single spikes and allows analysis of onset and propagation.