基于腹电信号的自动胎儿诊断算法的研究

Study of automatic fetal diagnostic algorithms based on abdominal signals

目的电子胎心监护是通过连续监测胎心率和宫缩的变化对胎儿宫内状态进行评估的重要手段,但对电子胎心监护的传统人工诊断存在主观性高和一致性低的局限。与超声多普勒法相比,经腹电信号法允许更长时间的监护,且受个体差异影响小,但国内缺乏精确的基于腹电信号的自动胎儿诊断算法。为此本研究提出了一种创新算法,旨在提高胎儿诊断的准确性和效率,并为国内临床医生提供有效的决策支持工具,提升医疗服务质量。方法本算法首先采用稳定有效的多任务深度学习网络对腹电信号提取到的胎心率进行分析得到胎心率参数(基线、加速和减速的起止时间),同时通过由腹电信号滤波得到的子宫肌电信号进行宫缩识别获取宫缩参数(宫缩频率和起止时间)。对提取的上述参数进行整理后结合电子胎心监护应用专家共识进行胎儿监护结果的诊断。结果通过分析89例同时采用腹电式动态胎儿监护仪和多普勒胎儿监护仪的20 min监护记录,算法在无激惹试验反应型的识别上表现出76.09%的敏感度和97.22%的阳性预测值,宫缩应激试验Ⅰ类的敏感度为88.89%,阳性预测值为68.09%,整体准确率达到77.53%。结论该算法在与医生诊断结果的对比中展现出较高的一致性,为提高临床决策质量和医疗服务提供了一个创新的辅助工具。

Objective Electronic fetal monitoring is an important means of assessing the intrauterine status of the fetus by continuously monitoring changes in fetal heart rate and uterine contractions.However traditional manual diagnosis of electronic fetal monitoring has the limitations of high subjectivity and low consistency.Compared to the ultrasound Doppler method,the transabdominal electrical signal method permits longer monitoring and is less affected by individual differences,yet there is a lack of accurate automated fetal diagnostic algorithms based on abdominal signals in China.For this reason,this study proposes an innovative algorithm that aims to improve the accuracy and efficiency of fetal diagnosis and to provide domestic clinicians with an effective decision-support tool to enhance the quality of healthcare services.Methods This algorithm uses a stable and effective multitask deep learning network to analyze the fetal heart rate extracted from the abdominal electrical signals to obtain the fetal heart rate parameters(baseline,acceleration and deceleration starting and stopping times),and the electrohysterogram signals obtained by filtering the abdominal signals are used for the recognition of uterine contractions to obtain the parameters of uterine contractions(uterine contraction frequency and starting and stopping times).These above-extracted parameters are sorted out and combined with the expert consensus on the application of electronic fetal monitoring for the diagnosis of fetal monitoring results.Results By analyzing the 20-minute monitoring recordings of 89 cases of simultaneous use of abdominal dynamic fetal monitor and Doppler fetal monitor,the algorithm demonstrates a sensitivity of 76.09% and a positive predictive value of 97.22% for the identification of none-stress test reactivity type,and a sensitivity of 88.89% and a positive predictive value of 68.09% for the contraction stress test category I,and an overall accuracy of 77.53%.Conclusions The algorithm demonstrates a high degree of consistency in comparisons with physicians' diagnostic results,providing an innovative aid for improving the quality of clinical decision-making and healthcare delivery.