てんかん脳波の棘波の自動検出について

概要

論文の詳細を見る
Automatic spike detection from epileptic EEG <Electroencephalogram> data is a pressing clinical problem in terms of locallization as well as quantification. In this study we obtained desirable results by using some parameters, to detect automatically all spikes from a long period recorded EEG by personal computer. Method : the EEG data are digitalized at 200 samples per second. (1) Compute some statistics : mean, variance, standard deviation (SD) , maximum and minimum in some fixed interval times. (2) Search absolute value of local maximum or local minimum larger than K_1・SD. (3) Compute left side duration time (NL) , right side duration time (NR) and peaked angle (TH) . (4) If (NL, NR < K_2 and NL+NR < K_3 and TH < K_4) then we define it as a spike, where K_1,K_2,K_3,K_4 are constants. (5) Time for each leaded channel of the spikes and informations : amplitude, convex or concave, etc. Result : 95% of the spikes detected by this method are real spikes, the remaining 5 % are sharp waves and noises. Conclusion : We can detect most of the spikes from the EEG. This definition is very simple and its BASIC program is too, furthermore the time to execute this program by a personal computer is considerably reduced.
聖徳大学の論文
1991-12-16