摘要
对正常听力20耳及蜗后病变14耳听性脑干反应用6阶AR模型参数进行分析,求出各阶最小均方差和(e0)及AR模型系数a1-a6.用多元逐步回归方法找出方差贡献较大的参数e0,a3和a6,并分别用Fisher准则及Bayes准则进行逐步判别分析,判别函数为:Y_1=0.8047 x_1-2.8044 x_2-3.9784,Y_2=0.1198x_1+1.7669x_2-1.7236(x_1=e0,x_2=a3),误判率为2.9％.结果表明用微机实现ABR自动诊断的可能性.
Parameters of a six -order autoregressive model were calculated , which were the sum of minimum mean square error (e0) and AR parameters from al to a6 for analysis of auditory brainstem responses (ABR) of 20 normal and 14 ictrocochlear lesion ears e0, a3 and a6 were parameters of the bigger contribution of variance, obtained by the method of stepwise regression analysis and stepwise discriminatory analysis was separatdy made according to Bayes criterion and Fisher criterion . The discriminating Junctions were Y1 =0.8047 x1 -2.8044 x2 -3.9784 , and Y2 =0.1198 X1+1.7669 x2-1.7236,The rate of discriminative error was 29% (1/34) .The results demonstrated the probability of auto - diagnosis of ABR by this method .
出处
《第四军医大学学报》
1991年第6期455-458,共4页
Journal of the Fourth Military Medical University
关键词
听性脑干反应
逐步判别分析
auditory brainstem response autoregressive model parameter, stcpwisc discriminant analysis
