Postpartum Hemorrhage Prediction Based on Convolutional Neural Networks

Postpartum hemorrhage (PPH) is one of the leading causes of maternal mortality and severe complications, particularly in low- and middle-income countries, where it has a higher incidence. Accurate prenatal prediction of PPH can help clinicians develop appropriate treatment strategies and reduce adverse pregnancy outcomes. Artificial intelligence (AI) has demonstrated great potential in disease diagnosis, largely due to the rapid advancements in deep learning (DL) technology. This study retrospectively analyzed MRI T2WI data from 280 pregnant women at Beijing Obstetrics and Gynecology Hospital, Capital Medical University. A convolutional neural network (CNN)-based model was proposed for PPH prediction, and a deep learning-Clinical Combined(DLC) model incorporating clinical factors was developed for PPH risk assessment. The model’s performance was evaluated using AUC, accuracy, and other metrics. The results showed that the DLC model achieved AUCs of 0.868 in the validation sets, respectively, outperforming the standalone DL model (AUC = 0.834) and the clinical model (AUC = 0.784). This study demonstrates that the DLC model, integrating imaging features and clinical factors, has high accuracy and clinical application potential for PPH prediction. It may provide valuable support for the early identification and prevention of PPH.