Developing Novel Models to Predict Fetal Autonomic Nervous System Maturation and Behavioural States

Abstract

Fetal brain development has lasting effects on an individual throughout their life; it is essential to identify and understand prenatal developmental disorders at an early stage. The neuro-vegetative control of the fetus plays a significant role in this regard. According to the concept of "fetal programming," stress during pregnancy has the potential to permanently alter the development of the fetal brain, resulting in disease later in life. Since heart rate is one of the non-invasive signals that can be obtained from a fetus, heart rate variability (HRV) analysis is uniquely appropriate for assessing the brain development of a fetus. However, even fetal autonomic development cannot be precisely assessed at this time. The vagal and sympathetic activity rhythms are traditionally considered to have different frequencies of heart rate fluctuations.

The HRV parameters increase fluctuation amplitude, increase complexity, and form patterns. These HRV indices may represent the development of the fetal autonomic control system. The development and emergence of fetal behavioral states are part of the maturation process, and it is essential to take these states into account as well. Previous studies have used threshold-based classification to identify fetal behavioral states (FBSes). In this study, FBSes were classified using the unsupervised k-means clustering algorithm, which divides n number of data into k number of clusters that need to be defined initially. Also, FBSes (Quiet state, Active state) were classified using different deep learning techniques (Bi-LSTM, CNN,CNN-Bi-LSTM).Two different inputs used are the heart rate time series 1D signal and its power spectral density of the low-frequency region. These inputs were compared to find the best one. The relationship between FBSes and the coupling of maternal and fetal heartbeats was analyzed. The thesis also discusses how maternal breathing patterns influence this coupling while considering FBSes. An autonomic interaction score (AIS) was developed considering maternal physiological factors, maternal demographics, and maternal-fetal heartbeats coupling together with fetal HRV parameters using the linear regression model and support vector regression model. It is concluded that AIS can be assessed based on universal developmental indices obtained from autonomic control patterns of the fetus and mother. AIS reflects normal complex functional brain maturation. Therefore, the findings presented in this study propose that the assessment of fetal development and health can be improved using AIS integrating fetal and maternal features.

Date of Award	10 May 2024
Original language	American English
Supervisor	AHSAN Khandoker (Supervisor)