Neonatal Sepsis Diagnosis Decision-Making Based on Artificial Neural Networks

Area of Science:

• Neonatal Medicine
• Artificial Intelligence in Healthcare
• Clinical Diagnostics

Background:

• Neonatal sepsis diagnosis is complicated by non-specific clinical presentations.
• Existing scoring systems lack individual patient consideration.
• Accurate early diagnosis is crucial for effective neonatal sepsis management.

Purpose of the Study:

• To develop an early and late-onset neonatal sepsis diagnosis model.
• To utilize clinical maternal and neonatal data from electronic health records.
• To create a predictive tool available at the time of clinical suspicion.

Main Methods:

• An artificial neural network (ANN) algorithm was trained and validated.
• A balanced dataset of septic and non-septic neonates (preterm and term) was used.
• The model incorporated 25 maternal and neonatal features from electronic records.

Main Results:

• The ANN model achieved 93.3% sensitivity and 80.0% specificity.
• The model demonstrated a 94.4% area under the receiver operating characteristic curve (AUROC).
• Performance surpassed physician diagnoses based on traditional scoring systems.

Conclusions:

• The developed ANN model offers a highly accurate method for neonatal sepsis diagnosis.
• Key predictive factors include maternal age, cervicovaginitis, and neonatal fever, apneas, and platelet counts.
• This AI-driven approach enhances diagnostic capabilities beyond conventional methods.