分层聚类分析方法确定α背景波的脑电地形图模式

Topographic classification of electroencephalographic alpha background activity patterns by hierarchical cluster analysis

目的:利用分层聚类模式识别方法对正常人α频段的单频率脑电地形图模式进行分类,确定正常人α频段的脑电地形图的模式。方法:选择1993-01/1998-12解放军总医院的研究生和工作人员为正常受试者214名(均经知情同意),年龄20～59岁。使用美国Neuroscan公司基于通用计算机系统的脑电采集系统,连续记录脑电信号。受试者在安静清醒条件下记录,使用10%～20%系统放置电极,单极记录28通道脑电图,参考电极为A1+A2。连续记录100～240s的脑电,取1s为1个时间段进行FFT变换,采样率256Hz,频率分辨率1Hz。用眼动伪迹矫正程序去除伪迹。使用脑电波幅频谱,脑电地形图用平均平方根功率(波幅频谱)表示。结果:214名受试者全部进入结果分析。①用分层聚类分析方法可以把α波脑电地形图归为12种模式。各个模式所占的百分比为:模式1:13.08%(28/214),模式2:28.22%(68/214),模式3:2.34%(5/214),模式4:0.93%(2/214),模式5:5.14%(11/214),模式6:1%(3/214),模式7:48%(16/214),模式8:3.27%(7/214),模式9:14.02%(30/214),模式10:5.14%(11/214),模式11:11.68%(25/214),模式12:3.74%(8/214)。②α波幅和分布明显影响α脑电地形图的模式。③根据聚类样本含量可以把α波脑电地形图归为5大模式,其中单一聚类的2个:聚类(1,n=28),聚类(2,n=68),复合聚类的3个:聚类(3+4+5+6+7,n=37),聚类(8+9,n=37),聚类(10+11+12,n=44)。④75.23%(161/214)的α波呈顶枕区分布,24.77%(53/214)的α波呈顶额区分布。结论:分层聚类分析是区别不同脑电地形图模式的有力工具。可以作为一个自动诊断工具使用。许多疾病可以产生不正常的脑电地形图模式,正常的α波脑电的模式分类和解释对脑电图临床诊断是非常重要的。

AIM： To identify the pattern of α rhythm in normal subjects by a topographic classifiaation of electroencephalographic （EEG） α background activity patterns with by hierarchical cluster analysis. METHODS： 214 normal subjects （informed and agreed） aged 20-59 years of postgraduate and staff of General Hospital of Chinese PLA from January 1993 and December 1998 were selected. Continuous EEG was collected with a PC-based EEG acquisition system （Neuroscan）. The examinees were recorded from 28-channels using 10%-20% system monopolor montage （AI＋A2） in a resting condition. The continuous 100-240 seconds EEG data were Segmented into 1 second epochs for FFT transform （sampled at 256 Hz, 1 Hz frequency resolution）. The epochs contaminated by eye movement-related artifacts were excluded from analysis using a rejection criterion procedure. With the EEG, the brain electrical activity map （BEAM） was expressed by root-mean-square power from amplitude spectrum data. RESULTS： All the 214 subjects were involved in the result analysis. ① The α topographical maps were classified into twelve patterns by Hierarchical cluster analysis. Percentage of the twelve different patterns： Pattern h 13.08% （28/214）, Pattern 2： 28.22% （68/214）, Pattern 3： 2.34% （5/214）, Pattern 4： 0.93% （2/214）, Pattern 5： 5.14% （11/214）, Pattern 6： 1% （3/214）,Pattern 7： 7.48% （16/214）,Pattern 8： 3.27% （7/214）, Pattern 9： 14.02% （30/214）, Pattern 10： 5.14% （11/214）, Pattern 11： 11.68% （25/214）, Pattern 12： 3.74% （8/214）. ②The amplitude and distribution of α activity obviously influenced classification of α topographic patterns. ③Five patterns were classified according to the quantity of subjects, including 2 single clusters： patterm 1=cluster （1, n=28）, pattern 2=cluster （2, n=68）, and 3 compound clusters： pattern 3=cluster （3＋4＋5＋6＋7, n=37）, pattern 4=cluster （8＋9, n=37）, pattern 5=cluster （10＋11＋12, n=44）. ④75.23% （161/214） α topography in parietal areas showed the occipital distribution and 24.77% （53/214） showed frontal distribution. CONCLUSION： Hierarchical cluster analysis is a very powerful method to diseriminate different EEG patterns. It ean be used as an automatie diagnostic tool. Different diseases can produce a number of abnormal EEG patterns and it is important to elassify and interpret normal α EEG patterns for clinical diagnosis.