Toward Intelligent Emergency Triage: A Feasibility Study of Real-Time Facial Expression-Based Chest Pain Intensity...

Area of Science:

• Artificial Intelligence
• Medical Technology
• Emergency Medicine

Background:

• Effective triage of chest pain patients in emergency settings is crucial but challenging, especially with communication barriers like face masks.
• Objective and rapid pain assessment is needed to support clinical decision-making in emergency departments.

Purpose of the Study:

• To develop and evaluate a real-time facial expression-based system for objective chest pain intensity assessment.
• To enhance emergency triage by providing rapid pain level recognition, particularly for non-communicative patients.

Main Methods:

• A YOLOv12-based facial expression recognition model was trained on annotated facial images of chest pain patients.
• The system categorizes pain into three levels: no pain, slight pain, and moderate to severe pain.
• System performance was evaluated using multiple YOLOv12 variants, with real-time and offline analysis modes.

Main Results:

• The YOLOv12-L model demonstrated high performance with 98.81% accuracy, 98.76% sensitivity, 98.79% specificity, 98.04% precision, and 98.41% F1-score.
• The system provides stable and accurate recognition of chest pain intensity from facial expressions.
• The system supports emergency triage by offering objective pain assessment for masked or non-communicative patients.

Conclusions:

• The developed facial expression-based system shows significant potential as a supportive tool for emergency triage workflows.
• Future research will focus on integrating edge computing for real-time pain assessment in ambulances to expedite patient intervention.