The Design of NEST: Neonatal Evaluation & Simulation Tool

Very preterm infants, born before 32 weeks of gestation, require intensive monitoring and support of their physiological functions in the Neonatal Intensive Care Unit (NICU). Due to their extreme fragility, monitoring options are limited, making timely and accurate clinical decision-making both critical and challenging for neonatologists.
To address these challenges, this project introduces the Neonatal Evaluation & Simulation Tool (NEST), a decision-support system that integrates patient data, machine learning algorithms, and physiology-based models into an intuitive, user-friendly interface.
After the system requirements were specified, a technical workflow was developed to guide data acquisition and model implementation. Two use cases were selected to evaluate the system. The first focused on predicting the failure of a second less invasive surfactant administration (LISA) treatment, which aimed at user interface evaluation. A logistic regression model was implemented using physiological and electronic medical record (EMR) data, alongside a cardiovascular-respiratory simulation model for educational purposes. The dashboard underwent two iterative development cycles, with user feedback informing improvements. Usability evaluation yielded a System Usability Scale (SUS) score of 80%, and requirement satisfaction scores exceeding 75%.
The second use case involved predicting the risk of developing late-onset sepsis (LOS) using a deep learning model, DeepLOS. The dashboard was adapted to visualize dynamic risk progression, and feedback from the model developer confirmed the clarity and effectiveness of the implementation workflow.
In summary, this project demonstrates a robust and adaptable workflow for integrating predictive and educational models into clinical dashboards. The proposed system enhances the accessibility of decision-support tools in neonatal care and establishes a foundation for future developments, including real-time data integration, model optimization and personalization, and regulatory compliance.