Sliding-window technique for the analysis of cerebral evoked potentials

To evaluate the efficiency of sliding-window technique in extracting and analyzing somatosensory evoked potentials (SEP) from multichannel electroencephalogram (EEG) data.

A time window of certain window size was moved along the time dimension of data sets. Values within the window were averaged for each trial, and then compared with a preset control window. The probability of randomly appeared significance resulting from repeated statistical comparison was calculated utilizing simulated EEG data sets. Cluster size (number of successive significant data points with given individual significance threshold) was determined to keep the general alpha value under 0.05. To test this procedure, multichannel EEG signals were recorded and analyzed from fourteen healthy right-handed volunteers, with painful and non-painful electrical stimuli delivered to the right middle fingers.

Cluster size increased in parallel with window size and individual statistical threshold. The major SEP components of real EEG data, as well as the difference between pain and non-pain SEPs, were demonstrated to be significant with the sliding-window method.

Sliding-window method is an effective tool for the analysis of SEP data.