Department of Perinatology, Ankara Etlik City Hospital, Ankara, Turkey.
Department of Obstetrics and Gynecology, Ankara Etlik City Hospital, Ankara, Turkey.
Am J Reprod Immunol. 2024 Oct;92(4):e70004. doi: 10.1111/aji.70004.
This study evaluates the association of novel inflammatory markers and Doppler parameters in late-onset FGR (fetal growth restriction), utilizing a machine-learning approach to enhance predictive accuracy.
A retrospective case-control study was conducted at the Department of Perinatology, Ministry of Health Etlik City Hospital, Ankara, from 2023 to 2024. The study included 240 patients between 32 and 37 weeks of gestation, divided equally between patients diagnosed with late-onset FGR and a control group. We focused on novel inflammatory markers-systemic immune-inflammation index (SII), systemic inflammatory response index (SIRI), and neutrophil-percentage-to-albumin ratio (NPAR)-and their correlation with Doppler parameters of umbilical and uterine arteries. Machine-learning algorithms were employed to analyze the data collected, including demographic, neonatal, and clinical parameters, to develop a predictive model for FGR.
The machine-learning model, specifically the Random Forest algorithm, effectively integrated the inflammatory markers with Doppler parameters to predict FGR. NPAR showed a significant correlation with FGR presence, providing a robust tool in the predictive model (Accuracy 77%, area under the curve [AUC] 0.851). In contrast, SII and SIRI, while useful, did not achieve the same level of predictive accuracy (Accuracy 75% AUC 0.818 and Accuracy 73% AUC 0.793, respectively). The model highlighted the potential of combining ultrasound measurements with inflammatory markers to improve diagnostic accuracy for late-onset FGR.
This study illustrates the efficacy of integrating machines with traditional diagnostic methods to enhance the prediction of late-onset FGR. Further research with a larger cohort is recommended to validate these findings and refine the predictive model, which could lead to improved clinical outcomes for affected pregnancies.
ClinicalTrials.gov identifier: NCT06372938.
本研究利用机器学习方法评估晚期胎儿生长受限（FGR）中新的炎症标志物和多普勒参数的相关性，以提高预测准确性。
这是一项 2023 年至 2024 年在安卡拉卫生部埃特利克城市医院围产医学系进行的回顾性病例对照研究。该研究纳入了 240 名妊娠 32 至 37 周的患者，平均分为晚期 FGR 患者组和对照组。我们重点关注新的炎症标志物——全身免疫炎症指数（SII）、全身炎症反应指数（SIRI）和中性粒细胞与白蛋白比值（NPAR）及其与脐动脉和子宫动脉多普勒参数的相关性。我们使用机器学习算法分析收集的数据，包括人口统计学、新生儿和临床参数，以建立 FGR 的预测模型。
机器学习模型（特别是随机森林算法）有效地将炎症标志物与多普勒参数相结合，以预测 FGR。NPAR 与 FGR 的存在显著相关，为预测模型提供了一个强大的工具（准确性 77%，曲线下面积[AUC]0.851）。相比之下，SII 和 SIRI 虽然有用，但没有达到相同的预测准确性（准确性分别为 75%，AUC 为 0.818 和准确性为 73%，AUC 为 0.793）。该模型强调了将超声测量与炎症标志物相结合以提高晚期 FGR 诊断准确性的潜力。
本研究表明，将机器与传统诊断方法相结合可以提高晚期 FGR 的预测效果。建议进行更大队列的研究以验证这些发现并改进预测模型，这可能会改善受影响妊娠的临床结局。
ClinicalTrials.gov 标识符：NCT06372938。
