Groome L J, Singh K P, Burgard S L, Collins H B, Bartolucci A A
Department of Obstetrics and Gynecology, University of Alabama at Birmingham, USA.
J Perinat Med. 1995;23(5):379-84. doi: 10.1515/jpme.1995.23.5.379.
The purpose of this study was to determine if discriminant analysis could be used to categorize fetal heart rate (FHR) - fetal eye movement (FEM) patterns. Statistical characteristics from 27 normal human fetuses at term for behavioral states, transitions, and insertions were established by combining the digitized FHR-FEM data for subjectively identical epochs. The mean FHR, the variance about the mean, and the presence or absence of FEM were calculated for each 3-min block in a sliding moving window with a 1-min step size. For each fetus, discriminant analysis was then used to assign 3-min blocks to either a behavioral state, a transition, or an insertion by comparing the statistical properties of a 3-min block with that of the data base. We found no difference between discriminant analysis and visual assignment in the average time spent in behavioral states 1F, 2F, and 4F, or in the mean duration of the transition/insertion periods. There was a highly significant linear relationship between computer-generated and visually-determined durations for behavioral states 1F (r = 0.972, p < 0.0001) and 2F (r = 0.989, p < 0.0001) and for the transition/insertion periods (r = 0.863, p < 0.0001). We conclude that discriminant analysis is a reliable computer-based method for behavioral state identification.
本研究的目的是确定判别分析是否可用于对胎儿心率（FHR）-胎儿眼动（FEM）模式进行分类。通过合并主观上相同时间段的数字化FHR-FEM数据，确定了27例足月正常胎儿在行为状态、过渡阶段和插入阶段的统计特征。在步长为1分钟的滑动窗口中，对每个3分钟时间段计算平均FHR、均值方差以及FEM的有无。然后，对于每个胎儿，通过将一个3分钟时间段的统计特性与数据库的统计特性进行比较，使用判别分析将3分钟时间段分配到行为状态、过渡阶段或插入阶段。我们发现，在行为状态1F、2F和4F所花费的平均时间以及过渡/插入阶段的平均持续时间方面，判别分析与视觉判定之间没有差异。对于行为状态1F（r = 0.972，p < 0.0001）和2F（r = 0.989，p < 0.0001）以及过渡/插入阶段（r = 0.863，p < 0.0001），计算机生成的持续时间与视觉判定的持续时间之间存在高度显著的线性关系。我们得出结论，判别分析是一种基于计算机的可靠行为状态识别方法。
