John E R, Prichep L S, Easton P
Brain Research Laboratories, New York University Medical Center, NY.
Psychiatry Res. 1994 Mar;55(1):13-40. doi: 10.1016/0925-4927(94)90009-4.
A large normative data base of visual and auditory event related potentials (ERPs) was collected. Factor analysis (PCVA) was used to extract factor wave shapes that accurately reconstructed these normal ERPs with appropriate "factor scores." The mean value and standard deviation (SD) of the normative factor score distribution were computed separately for each stimulus, factor, and electrode. This enabled reconstruction of any individual ERP as a combination of these standardized Varimax descriptors, with z-transformation of the required factor scores giving objective statistical assessment of ERP wave shape. Statistical probability factor z-score topographic maps were constructed, color coded in SDs from the normative means. The incidence of significant individual deviations from these normative mean values was at or near chance levels in test groups of normal subjects. For many of these new ERP descriptors, significant deviations from the norms were found for single features in from 20% to as much as 63% of the patients in particular diagnostic categories. Factor z-scores were used to construct multivariate discriminant functions that accurately and replicably separated (1) normal from schizophrenic from demented subjects and (2) schizophrenic from bipolar depressed subjects.
收集了一个关于视觉和听觉事件相关电位（ERP）的大型标准化数据库。使用因子分析（主成分方差分析）来提取能够通过适当的“因子得分”准确重构这些正常ERP的因子波形。分别针对每个刺激、因子和电极计算标准化因子得分分布的均值和标准差（SD）。这使得能够将任何个体ERP重构为由这些标准化的方差最大化描述符组成的组合，对所需因子得分进行z变换可对ERP波形进行客观的统计评估。构建了统计概率因子z得分地形图，根据与标准化均值的标准差进行颜色编码。在正常受试者测试组中，个体与这些标准化均值存在显著偏差的发生率处于或接近随机水平。对于许多这些新的ERP描述符，在特定诊断类别的患者中，有20%至高达63%的患者在单个特征上与规范存在显著偏差。因子z得分用于构建多变量判别函数，该函数能够准确且可重复地将（1）正常受试者与精神分裂症患者和痴呆患者区分开来，以及（2）精神分裂症患者与双相抑郁症患者区分开来。
