Automated Pain Assessment in Children Using Electrodermal Activity and Video Data Fusion via Machine Learning.

Area of Science:

• Biomedical Engineering
• Pain Medicine
• Pediatrics

Background:

• Assessing pain in children, especially nonverbal ones, presents significant challenges.
• Current methods rely heavily on subjective self-reports, which can be unreliable.
• Objective pain metrics are needed to supplement existing assessment techniques.

Purpose of the Study:

• To develop and validate an objective pain assessment metric for children.
• To fuse electrodermal activity (EDA) and video facial expression data for enhanced pain detection.
• To evaluate the efficacy of this multimodal approach in post-operative pediatric patients.

Main Methods:

• Collected electrodermal activity (EDA) and video facial expression data from children post-laparoscopic appendectomy.
• Utilized a weighted maximum likelihood algorithm for feature selection from EDA and facial expression data.
• Developed an automated classification algorithm to distinguish clinically significant pain from non-significant pain.

Main Results:

• Individually, EDA and facial expression data showed above-chance accuracy in predicting pain.
• Fusion of EDA and facial expression data significantly improved pain classification accuracy.
• The combined approach achieved 90.91% accuracy, 100% sensitivity, and 81.82% specificity for clinically significant pain.

Conclusions:

• Multimodal objective pain assessment using EDA and facial expressions is effective in children.
• This approach provides a reliable, nonverbal method for pain evaluation in pediatric post-operative recovery.
• The developed algorithm offers an accurate and objective tool for clinical pain assessment in children.