An Explainable Fuzzy Framework for Assessing Preeclampsia Classification.

Area of Science:

• Medical Informatics
• Computational Biology
• Maternal Health

Background:

• Preeclampsia is a major global cause of maternal morbidity.
• Accurate and interpretable predictive systems are crucial for clinical decision-making.
• Existing models often lack transparency, hindering clinical adoption.

Purpose of the Study:

• To introduce SK-MOEFS, an explainable framework for preeclampsia risk classification.
• To develop a system that generates clinically interpretable rules.
• To enhance decision-making in maternal care through transparent AI.

Main Methods:

• Utilized fuzzy decision trees integrated with a genetic algorithm for rule optimization.
• Employed multi-objective evolutionary optimization for accuracy and interpretability.
• Trained and validated on a multi-ethnic cohort of 574 third-trimester pregnancies using open-source tools.

Main Results:

• Achieved 91% classification accuracy, 0.89 AUC, and 0.88 recall.
• Outperformed standard interpretable models in predictive accuracy and transparency.
• Demonstrated a focus on minimizing false negatives for critical risk stratification.

Conclusions:

• SK-MOEFS offers a transparent and reliable approach to preeclampsia risk prediction.
• The framework's rule translation and natural language output enhance clinical utility.
• Provides a bridge between AI and clinical judgment for improved maternal care outcomes.