1998 年 36 巻 3 号 p. 200-209
Spikes detected on EEG (electroencephalogram) records give important information for clinical diagnosis, especially that of epileptic disorders, and therefore automatization of spike detection will be a powerful aid for electroencephalographer in the visual inspection of spikes. A new method for detecting spikes in EEG records was developed by taking account of properties of the spikes which had been employed by electroencephalographers for the visual inspection. The proposed method has two key features as follows: 1) adjustment of the threshold values for the spike detection depending upon the characteristics of spikes in each EEG record and 2) renovation of the templates for the spike waveform according to the detected spikes. The discrimination of the artifacts such as eye blinks and EMG (electromyogram) from the spikes was also taken into account. The proposed method was applied to 71 segments (1 segment=5sec) of EEGs recorded from 10 patients. A detection accuracy of 85.3% (64/75) was obtained as the result of the automatic detection, which was in good agreement with the electroencephalographer's visual inspection. Since the proposed method was based on adaptive decision criteria, it is automatically adaptable to the spike of various waveforms derived from respective subjects. The proposed spike detection method can be effectively applicable for clinical use.