Automated Electrodermal Activity and Facial Expression Analysis for Continuous Pain Intensity Monitoring on the X-ITE...

Area of Science:

• Biomedical Engineering
• Healthcare Technology
• Pain Management

Background:

• Continuous patient pain monitoring is crucial for healthcare quality.
• Electrodermal Activity (EDA) and facial expressions are key indicators of pain intensity.
• Previous studies have explored these modalities separately.

Purpose of the Study:

• To develop and evaluate an automated system for continuous patient pain intensity monitoring.
• To compare the effectiveness of Electrodermal Activity (EDA) and facial expression analysis.
• To investigate the benefits of late fusion of EDA and facial expression data.

Main Methods:

• Utilized Electrodermal Activity (EDA) sensors and facial expression analysis.
• Applied machine learning models: Random Forest (RF), Long-Short Term Memory Network (LSTM), and LSTM with sample-weighting (LSTM-SW).
• Employed late fusion techniques to combine data from both modalities.

Main Results:

• Late fusion of EDA and facial expressions demonstrated effectiveness in balanced datasets (Micro F1-score ~61%, ICC ~0.35).
• LSTM and LSTM-SW models showed superior performance for EDA regression in imbalanced datasets.
• Both integrated and EDA-only approaches outperformed baseline and random guessing methods.

Conclusions:

• Integrating EDA and facial expression modalities provides valuable insights into patient pain.
• EDA-based regression models, particularly LSTM variants, are effective for pain monitoring, especially in imbalanced scenarios.
• The developed system can enhance medical centers' ability to assess and respond to patient pain.