MAbortionPre: Predicting the Risk of Missed Abortion through Complete Observation of Clinical Data

During pregnancy, childbirth, and postpartum, pregnant women may encounter various physical issues, one of which is missed abortion (also known as delayed abortion). Its causes are complex and diverse, including genetic, endocrine, immune, and other factors. Accurate prediction of the risk of missed abortion is crucial for the health of pregnant women and fetuses. To address this issue, we have collected and analyzed clinical data of pregnant women, including age, laboratory test results, etc. We have established a machine learning model called MAbortionPre to predict the risk of missed abortion. Considering statistical analysis, we have designed a missing value imputation module using data with exploitable patterns, and utilized covariance structures to ensure data integrity and interpretability. In addition, we have focused on designing a method for selecting data features, utilizing weighted Lasso to select different feature combination for analysis, this ensures that the model achieves optimal results while avoiding overfitting to noise features. We compared our model with the baseline, and the experimental results showed that our model has high accuracy and stability, which can help doctors intervene at an early stage to reduce the incidence of missed abortion, demonstrating significant clinical significance.