Noise Reduction for Detecting Event-Related Potential by Processing in Dipole Space

概要

論文の詳細を見る
Averaged responses are generally used to detect event-related potentials (ERPs) by supressing the background electroencephalography (EEG) wave, the ERP component of a single-trial response, or the average of a small number of responses is used to assess time variation in a subjects’ state in detail. We therefore proposed a new method of reducing the noise component including the background wave in a single-trial response. In this study, our target is a component such as $N_{100}$ approximated by one dipole. The method was performed by modifying the dipole position in the brain and detecting the projected components with reference to the dipole estimated from an averaged response. Results of simulation indicate that the proposed method could improve signal-to-noise ratio by 7.6 dB and decrease the error in $N_{100}$ peak latency 6.7 ms by suppressing the influence of the background wave. In the EEG experiment, eight healthy subjects paticipated and their results show that the sway of waveforms by the background wave is suppressed and that the peak of the $N_{100}$ component becomes prominent compared with that of the original single-trial response.
2007-06-15