Abstract
Detection and extraction of a single waveform, which appears in electroencephalogram (EEG) data, are an important step for neurophysiological diagnosis. The correlation coefficient is a useful index for correlating two similar signals of variable amplitude, but does not apply to signals of variable duration. In this study, a similarity coefficient was defined and used for extracting the single waveform of both variable duration and variable amplitude. The similarity coefficient was defined by introducing a time-scale factor into the correlation coefficient to measure the similarity of two signals. A single waveform of variable duration and variable amplitude in an EEG record was accurately detected and extracted by the use of the similarity coefficient. The method was evaluated using the simulated data of a single waveform with noise and then applied to actual EEG data contaminated with epileptic spikes. The spikes were accurately detected and extracted from the raw EEG data using the similarity coefficient. Furthermore, the spikes were subtracted from the raw EEG to make the background EEG stand out. The proposed method, with the similarity coefficient, is widely applicable for extracting any kind of burst signal of EEG record and evoked potentials for extracting the characteristics of single waveforms.
