Early Prediction of Preeclampsia Using a Multimodal Machine Learning Model Integrating Clinical, Hemodynamic, and Biochemical Data

Cardenas-Urrelo R.; Quinteros-Navarro D.; Tang C.; Torres-Flores Y.; Navarro Y.; Soto M.; Maguiña A.; Huamantumba M. · Ingenierie des Systemes d'Information (2026) · DOI: 10.18280/isi.310328

Resumen

Preeclampsia (PE) remains a major cause of maternal and perinatal morbidity and mortality, and its early prediction continues to be a significant clinical challenge. This study develops a multimodal machine learning (ML) model that integrates clinical, hemodynamic, and biochemical variables collected during the first trimester to improve early risk stratification for PE. We analyzed a cohort of 1,000 pregnant women, incorporating over 40 predictors, including maternal characteristics, mean arterial pressure (MAP), uterine artery pulsatility index (UtA-PI), and placental biomarkers. The model demonstrated strong predictive power with an area under the ROC curve (AUC) of 0.85, achieving a sensitivity of 52% and specificity of 94%. The key predictors identified were MAP, UtA-PI, and placental growth factor (PlGF), with Precision-Recall analysis confirming robust performance despite class imbalance. The results indicate that this model can serve as a reliable first-trimester screening tool for PE, offering the potential for integration into routine prenatal care to identify high-risk pregnancies early. However, further validation in multicenter studies and diverse populations is necessary to assess the model’s generalizability, stability, and clinical applicability. This research presents a promising step toward personalized prenatal care by providing a data-driven tool to support decision-making in the early detection of preeclampsia.

Instituciones: Universidad Tecnológica del Perú; Universidad Nacional Federico Villarreal; Universidad Nacional de San Martín; Universidad Privada del Norte