Early Detection of Preeclampsia and Gestational Hypertension Using Machine Learning Techniques

Abstract

Gestational hypertension is a maternal entanglement, manifested through increased blood pressure which may occur later than week 20 of gestation period and may result in severe snags viz. preeclampsia, preterm neonatal delivery and injury of obstetric organs. Early and precise prediction is important because early tracking and cure would protect the pregnant women and the fetus from great number of dangers. Computational classification models are crucial for forecasting hypertensive disorders in pregnant women with analysis of large volumes of clinical, demographic and physiological data. Among the different predictive characteristics, Systolic Blood Pressure (SBP) and Diastolic Blood Pressure (DBP) are of particular concern, since the aberrant changes of SBP and DBP are considered to be directly related to the presence of gestational hypertension and preeclampsia. The classification models that have been employed in this study in the analysis are the Mutual Information based classification, Recursive feature elimination (RFE), Lasso Regression based classification and Tree based classification. Moreover, Hybrid Feature Selection strategy has been suggested which is a mixture of the Filter Method (Mutual Information) and an Embedded Tree-Based Model, which enhances the predictive accuracy. All the models were measured by the important characteristics like F1-score, precision, accuracy, specificity, and sensitivity. The findings show that the suggested hybrid method was the most successful and recorded the best accuracy of 98% in all the datasets and an acceptable result in the training and testing data. The paper represents various machine learning classification algorithms to forecast gestational hypertension and preeclampsia. These are optimized models that can help in diagnosing the issues at an early stage and also improve the results of the maternal healthcare by taking into consideration the variables of SBP, DBP and other clinically significant variables.