Modelling the risk of abnormal pregnancy-related outcomes such as stillbirth and preterm birth have been proposed in the past. Commonly they utilize maternal demographic and medical history information as predictors, and they are based on conventional statistical modelling techniques. In this study, we utilize state-of-the-art machine learning methods in the task of predicting early stillbirth, late stillbirth and preterm birth pregnancies. The aim of this experimentation is to discover novel risk models that could be utilized in a clinical setting. A CDC data set of almost sixteen million observations was used conduct feature selection, parameter optimization and verification of proposed models. An additional NYC data set was used for external validation. Algorithms such as logistic regression, artificial neural network and gradient boosting decision tree were used to construct individual classifiers. Ensemble learning strategies of these classifiers were also experimented with. The best performing machine learning models achieved 0.76 AUC for early stillbirth, 0.63 for late stillbirth and 0.64 for preterm birth while using a external NYC test data. The repeatable performance of our models demonstrates robustness that is required in this context. Our proposed novel models provide a solid foundation for risk prediction and could be further improved with the addition of biochemical and/or biophysical markers.

No abstract

Objective Minority oversampling is a standard approach used for adjusting the ratio between the classes on imbalanced data. However, established methods often provide modest improvements in classification performance when applied to data with extremely imbalanced class distribution and to mixed-type data. This is usual for vital statistics data, in which the outcome incidence dictates the amount of positive observations. In this article, we developed a novel neural network-based oversampling method called actGAN (activation-specific generative adversarial network) that can derive useful synthetic observations in terms of increasing prediction performance in this context. Materials and Methods From vital statistics data, the outcome of early stillbirth was chosen to be predicted based on demographics, pregnancy history, and infections. The data contained 363 560 live births and 139 early stillbirths, resulting in class imbalance of 99.96% and 0.04%. The hyperparameters of actGAN and a baseline method SMOTE-NC (Synthetic Minority Over-sampling Technique-Nominal Continuous) were tuned with Bayesian optimization, and both were compared against a cost-sensitive learning-only approach. Results While SMOTE-NC provided mixed results, actGAN was able to improve true positive rate at a clinically significant false positive rate and area under the curve from the receiver-operating characteristic curve consistently. Discussion Including an activation-specific output layer to a generator network of actGAN enables the addition of information about the underlying data structure, which overperforms the nominal mechanism of SMOTE-NC. Conclusions actGAN provides an improvement to the prediction performance for our learning task. Our developed method could be applied to other mixed-type data prediction tasks that are known to be afflicted by class imbalance and limited data availability.

This diagnostic study assesses the use of dried blood samples to detect prohormone brain-type natriuretic peptide for diagnosis of congenital heart disease in neonates in Sweden.

No abstract