Recovery of Signal Loss in Fetal Heart Rate Dynamics with Residual Bootstrap

Abstract

A non-stress test (NST) is used to evaluate the fetal condition through monitoring the patterns and variabilities of fetal heart rate (FHR) by the effects of the autonomic nervous system (ANS). This is the most common prenatal test in a noninvasive manner. FHR data is presented in a type of time series, which is accompanied with signal loss due to physical and biological causes. This limits the diagnosis of fetal growth and well-being. In this thesis, we propose a method to recover/estimate the FHR data with high rate of signal loss by applying the time series model and residual bootstrap method. In order to reduce the bias in signal loss estimation, the residuals obtained from the predicted time series model are applied to the bootstrap, a representative nonparametric method. The validity of the proposed method is verified by the nonlinear dynamic indices reflecting the chaotic nature of FHR and the meaningful outcomes of the prediction of fetal health and the classification of high-risk pregnancy are produced.| 자율신경계에 의한 태아심박동수의 증감과 박동간 변이를 관찰하는 비수축 검사는 산모에게 자극을 주지 않으면서 태아의 상태를 평가할 수 있는 가장 일반적인 산전검진법이다. 이러한 태아심박동 자료는 시계열의 형태로 나타나는데, 물리, 생물학적 원인으로 인하여 태아심박동에 신호 손실이 발생하고, 이는 태아의 성장과 안녕 진단에 제약을 준다. 본 논문에서는 손실률이 큰 태아심박 자료를 시계열 모형과 잔차 붓스트랩 방법을 적용하여 복구/추정하는 방법을 제안하였다. 예측 시계열 모형로부터 얻어진 잔차를 비모수적 방법인 붓스트랩에 적용함으로써 신호 손실 추정에서의 편의(bias)을 줄이고자 하였다. 제안된 방법은 심박동의 카오틱한 특성을 반영하는 비선형 지수들에 의해 그 효용성이 확인되었으며 태아 건강 상태의 예측과 고위험 임신의 분류에 유의한 결과를 보여주었다.; A non-stress test (NST) is used to evaluate the fetal condition through monitoring the patterns and variabilities of fetal heart rate (FHR) by the effects of the autonomic nervous system (ANS). This is the most common prenatal test in a noninvasive manner. FHR data is presented in a type of time series, which is accompanied with signal loss due to physical and biological causes. This limits the diagnosis of fetal growth and well-being. In this thesis, we propose a method to recover/estimate the FHR data with high rate of signal loss by applying the time series model and residual bootstrap method. In order to reduce the bias in signal loss estimation, the residuals obtained from the predicted time series model are applied to the bootstrap, a representative nonparametric method. The validity of the proposed method is verified by the nonlinear dynamic indices reflecting the chaotic nature of FHR and the meaningful outcomes of the prediction of fetal health and the classification of high-risk pregnancy are produced.