Prediction for the development of preeclampsia through non-invasive hemodynamics using machine learning, distinguishing early from late

Preeclampsia (PE) and hypertensive disorders of pregnancy (HDP) are major contributors to maternal-fetal morbidity and prematurity worldwide. These conditions are classified as early- or late-onset based on gestational timing. This study investigates the potential of impedance cardiography (ICG) as a tool for non-invasive hemodynamic assessment to predict early versus late-onset PE risk. Using machine learning techniques, specifically the J48 classification tree algorithm, a predictive model was developed to identify novel hemodynamic patterns beyond conventional metrics. A total of 405 high-risk pregnant patients between 17 and 33 weeks of gestation were evaluated, with hemodynamic parameters assessed using ICG. The study aimed to differentiate between early-onset and late-onset PE and to explore non-traditional hemodynamic variables associated with its development.

Results demonstrated that the machine learning model accurately identified high-risk pregnant women who developed PE, achieving a 95% correct classification rate. Furthermore, the model effectively distinguished between early- and late-onset cases. Importantly, it incorporated variables related to contractility, cardiovascular performance, and afterload, underscoring the potential of non-invasive hemodynamic assessment for early PE detection.

Despite certain limitations, including the modest number of PE events and the necessity for external validation, these findings highlight the promise of artificial intelligence in improving risk prediction and advancing clinical management strategies for PE in high-risk pregnancies.