Premature infants require individualized nutritional management to support optimal growth and neurodevelopment, yet adherence to neonatal nutrition guidelines remains challenging in busy intensive care units.
This research introduces the Nutrition Recommendation Ontology (NRO), an ontology- and rule-based clinical decision support system that integrates electronic medical record data with standardized neonatal nutrition guidelines. The system models neonatal clinical information using semantic technologies, knowledge graphs, and logical reasoning to generate personalized recommendations for enteral and parenteral nutrition while automatically identifying deviations from established guidelines.
The framework was developed using retrospective data from 601 NICU patients representing 8,460 patient-days of clinical care. The ontology consists of 121 classes, 366 axioms, and 157 semantic rules, enabling automated reasoning based on gestational age, day of life, clinical conditions, risk factors, and neonatal morbidities.
Validation by an expert panel of neonatologists demonstrated 98% agreement between the system-generated recommendations and expert clinical decisions, highlighting the promise of semantic artificial intelligence for improving nutritional management and supporting evidence-based decision making in neonatal intensive care units.