Enhancing care strategies for preterm pregnancies by using a prediction machine to aid clinical care decisions

Preterm births are one of the main causes of death in children under the age of 5; they carry financial implications to the economy and cause exceptional psychological distress to mothers and their families. Prior work has been done in the application of classification methods towards predicting whether a pregnant patient is likely to deliver preterm, but the majority of these solutions have been done separately without full consideration of how their proposed solution can be integrated into a clinical system setup. In this work, we propose a multi order cybernetic framework to design a recommender system that can be used to form a closed loop clinical interaction poised towards steering a system from its current state into a more desirable outcome. Using fused estimates from both electrohysterogram (electrophysiology) and tocogram (mechanical) signals from uterine wall contractions, a classification machine was designed to classify between Preterm/Term states, and also predict an associated delivery imminency for the pregnant patient. The classification machine was implemented using a multilayer perceptron neural network (MLP) and a support vector machine (SVM), where it was seen that the SVM outperformed the MLP in the majority of the classification tasks. Further work in this area would involve the application of regression techniques towards the classification tasks, which is expected to also provide greater model interpretability and continuous state estimation.