Children's Pain Identification Based on Skin Potential Signal

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Pain Management Research

Background:

• Pain assessment in children is challenging due to communication barriers.
• Current subjective pain scales have limitations in objectivity and reliability.
• Developing objective pain indicators is crucial for effective pediatric pain management.

Purpose of the Study:

• To develop and validate a novel pain assessment scheme using skin potential signals.
• To convert subjective pain experiences into objective, machine-learnable indicators.
• To explore the efficacy of machine learning algorithms for automated pain identification.

Main Methods:

• Designed a portable, non-invasive device to measure skin potential signals.
• Collected data from 623 subjects, with 358 valid records (218 silent, 262 pain samples).
• Extracted 38 features, identified 7 key features, and applied random forest classification.

Main Results:

• The random forest algorithm achieved a pain identification accuracy of 70.63%.
• Seven specific features demonstrated superior performance in distinguishing pain from non-pain states.
• Results indicate potential for objective pain assessment, though current accuracy is below state-of-the-art (81.5%).

Conclusions:

• The proposed skin potential-based pain assessment scheme shows promise for objective pain identification in clinical settings.
• Clinical pain stimuli, induced by operations, present challenges in controlling intensity, impacting accuracy.
• Further refinement of the methodology is warranted to enhance accuracy and clinical utility for pain management.