Designing an Intelligent Nursing Decision Support System Based on Artificial Intelligence for Monitoring Preterm Infants in the Neonatal Intensive Care Unit (NICU)

Introduction: Preterm infants require continuous and accurate monitoring in the Neonatal Intensive Care Unit due to the immaturity of their vital systems. The complexity and dynamism of these infants' clinical conditions make early detection of complications such as sepsis, apnea, and hemodynamic fluctuations challenging. Objective: This study aimed to design and simulate an intelligent nursing decision support system based on hybrid deep learning for simultaneous monitoring of vital variables in preterm infants. Methods: This developmental-applied study used simulated data from ۱۰۰۰ preterm infants generated using multivariate statistical models and valid physiological relationships. Five main variables including heart rate, respiratory rate, arterial oxygen saturation, body temperature, and arterial blood pressure were simulated at ۳۰-second intervals for ۹۶ hours. A hybrid deep model comprising Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) with three hidden layers was designed and trained on ۷۰% of the data. Model performance was evaluated using accuracy, sensitivity, specificity, F۱-Score, and Area Under the ROC Curve. Results: The designed hybrid CNN-LSTM model identified clinical instability and patterns leading to critical events with an accuracy of ۹۴.۷%, sensitivity of ۹۳.۲%, specificity of ۹۵.۸%, F۱-Score of ۰.۹۴, and AUC of ۰.۹۸. The mean early prediction time of critical events was ۱۸.۷ minutes before the onset of overt clinical symptoms, reaching ۳۱.۵ minutes for sepsis. The false alarm rate in the proposed model (۴.۲%) was significantly reduced compared to the traditional method (۲۵.۵%). Conclusion: The intelligent decision support system based on hybrid CNN-LSTM architecture, with the capability of simultaneous extraction of spatial and temporal features, has high potential for early identification of clinical instability in preterm infants. This system can serve as a supportive tool alongside NICU nurses, reducing cognitive load and false alarms while enhancing care safety.

کلیدواژه ها:

Artificial Intelligence ، Clinical Decision Support Systems ، Preterm Infant ، Neonatal Intensive Care Unit ، Deep Learning ، CNN-LSTM ، Early Warning ، Alarm Fatigue