Summary of the extracted features.

BackgroundChorioamnionitis is recognized as a major consequence of preterm premature rupture of membranes (PPROM), and a frequent cause of neonatal morbidity and mortality. The association between fetal heart rate (FHR) and chorioamnionitis remains unclear.ObjectivesThe aim of this study was to evaluate the dynamics of FHR in a PPROM population at the approach of delivery according to the presence or absence of chorioamnionitis.Materials & methods120 pregnant women with PPROM between 26 and 34 weeks’ gestation were enrolled in this multicenter prospective unblinded study. All participants were fully informed of the study’s objectives. 39 of the 120 patients were included in the analysis of FHR recordings. The analysis consisted of extracting features from computerized FHR analysis (cFHR) and fetal heart rate variability analysis (FHRV) in the temporal, frequency and nonlinear domains. Then, each set of features was analyzed separately using the multiple factor analysis, where three groups were defined as the feature set for days 0, -1 and -2 prior to birth. The distances between the global projection and the projections for each day were computed and used in the ROC analysis to distinguish chorioamnionitis from non-chorioamnionitis group.ResultsThe results showed that there were significant differences in certain features between populations with and without chorioamnionitis. The distinction between the two populations reached an area under the curve (AUC) of only 37% [34–40] for cFHR features and 63% [59–66] for time-domain FHRV features when comparing all stages of chorioamnionitis to non-chorioamnionitis subjects. When only stage 3 chorioamnionitis was compared to non-chorioamnionitis patients, the AUC reached 90% [88–93] for nonlinear-domain and 84% [82–87] for time-domain FHRV features, whereas it was limited to 71% [68–74] using cFHR features.ConclusionThe present study suggests that the HRV features are more reliable for diagnosing chorioamnionitis than cFHR, and that the assessment of features dynamics over several days is an interesting tool for detecting chorioamnionitis. Further study should be carried out on a larger sample to confirm these findings, improve the diagnostic performance of chorioamnionitis and help clinicians decide on delivery criteria.

背景：绒毛膜羊膜炎（Chorioamnionitis）是早产胎膜早破（preterm premature rupture of membranes, PPROM）的主要并发症，亦是新生儿发病与死亡的常见病因。胎儿心率（fetal heart rate, FHR）与绒毛膜羊膜炎之间的关联尚未明确。
目的：本研究旨在针对早产胎膜早破人群，依据绒毛膜羊膜炎的有无，分析分娩临近阶段的胎儿心率动态变化特征。
材料与方法：本研究为多中心前瞻性非盲试验，共纳入120名孕26~34周的早产胎膜早破孕妇，所有受试者均充分知悉本研究目的。120名受试者中，共39名被纳入胎儿心率记录分析。分析流程包括从计算机辅助胎儿心率分析（computerized FHR, cFHR）及时域、频域与非线性域的胎儿心率变异性（fetal heart rate variability, FHRV）分析中提取特征。随后，针对分娩前0天、-1天及-2天的特征集，分别采用多因素分析进行处理。计算全局投影与每日投影间的距离，并将其用于受试者工作特征（ROC）分析，以区分绒毛膜羊膜炎患者与非患者组。
结果：研究显示，绒毛膜羊膜炎阳性与阴性人群的部分特征存在显著差异。当对比所有分期的绒毛膜羊膜炎患者与非患者时，计算机辅助胎儿心率特征的受试者工作特征曲线下面积（AUC）仅为37%[34~40]，时域胎儿心率变异性特征的AUC为63%[59~66]。若仅对比3期绒毛膜羊膜炎患者与非患者，非线性域胎儿心率变异性特征的AUC可达90%[88~93]，时域胎儿心率变异性特征的AUC为84%[82~87]，而计算机辅助胎儿心率特征的AUC仅为71%[68~74]。
结论：本研究表明，胎儿心率变异性特征相较于计算机辅助胎儿心率特征，在诊断绒毛膜羊膜炎时具备更高的可靠性；且连续数日的特征动态评估可作为检测绒毛膜羊膜炎的有效工具。未来需开展更大样本量的研究以验证本研究结论，进一步提升绒毛膜羊膜炎的诊断效能，辅助临床医师制定分娩决策标准。