Dynamic gestational week prediction model for pre-eclampsia based on ID3 algorithm.

Area of Science:

• Obstetrics and Gynecology
• Medical Informatics
• Computational Biology

Background:

• Pre-eclampsia (PE) is a serious pregnancy complication characterized by hypertension and placental dysfunction.
• It poses significant risks to both maternal and fetal health, stemming from placental damage, ischemia, hypoxia, and oxidative stress.
• Identifying risk factors and predicting PE onset are crucial for timely intervention.

Purpose of the Study:

• To develop and evaluate a machine learning model for predicting pre-eclampsia (PE).
• To incorporate clinical risk factors and gestational week for improved PE prediction.
• To differentiate between early-onset pre-eclampsia (EOPE) and late-onset pre-eclampsia (LOPE).

Main Methods:

• Utilized hospital medical record data to identify risk factors across clinical epidemiology, hemodynamics, biochemistry, and biomarkers.
• Applied the decision tree ID3 algorithm for developing a dynamic gestational week prediction model for PE.
• Screened and analyzed model parameters in subgroups: early-onset pre-eclampsia (EOPE) and late-onset pre-eclampsia (LOPE).

Main Results:

• The machine learning model achieved an overall accuracy of 86% for PE prediction.
• Macro-average performance metrics were: precision = 76%, recall = 73%, F1-score = 75%.
• Weighted-average performance metrics demonstrated high effectiveness: precision = 88%, recall = 89%, F1-score = 89%.

Conclusions:

• The inclusion of gestational week as a dynamic parameter significantly enhances the clinical applicability of the PE prediction model.
• The developed model effectively predicts pre-eclampsia, including its early-onset and late-onset subtypes.
• This approach offers a convenient and effective tool for clinical risk assessment and prediction of pre-eclampsia.