A predictive model for the risk of fetal and maternal injury during childbirth

Thesis

Vaginal birth requires the large deformation of both maternal and fetal tissues, with a risk of injury to both parties. Understanding the loading experienced by the maternal pelvic floor and fetal head as the fetus travels through the birth canal would bring insight into the mechanisms of such injuries. Furthermore, being able to predict these loads and translate them into clinical outcomes ahead of labour has the potential to help guide the clinician in their decision-making regarding a patient’s birth and improve fetal and maternal outcomes altogether.

This thesis aims to develop and validate a predictive model of the second stage of labour, that incorporates both mechanical and patient-specific information, to predict the vaginal labour outcomes for a given patient’s birth.

Parametric finite element models of both mother and fetus were developed, parametrised with ultrasound measurements. These models were used to create a finite element library of labour simulations, through the variation of maternal and fetal model parameters and two different fetal positions. This labour simulation library was then used to train a more cost effective machine learning surrogate model to predict relevant mechanical metrics. A clinical labour outcome dataset, provided by Myers soft tissue lab from a single tertiary care centre in New York, New York, USA, was coupled with the results from the surrogate model to train a machine learning based classifier, able to predict clinical labour outcomes for a given patient.

The finite element simulations of labour agreed with clinical measurements and trends, while also agreeing with existing finite element simulations of labour in the literature. The fetal head model was able to provide insight into the loading of the fetal brain during labour, for a range of labour scenarios. Additionally, fetal position was found be a key feature in prediction the majority of mechanical metrics during labour. Finally, the pipeline for predicting clinically relevant labour outcomes using mechanics informed machine learning, showed potential for predicting injury in a low-risk cohort.

This work provides a framework for a mechanics informed predictive model for clinical labour outcomes. While at the proof-of-concept stage, this demonstrates the potential to inform clinical decision-making and improve patient outcomes during labour.

Authors: Collier, A

• DOI: 10.5287/ora-o0bwp6dmy
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: finite element simulations; fetal head moulding; vaginal labour; childbirth; machine learning
• Subjects: Labor (Obstetrics); Finite element method; Biomedical engineering