Pain Assessment Tool With Electrodermal Activity for Postoperative Patients: Method Validation Study.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Machine Learning in Healthcare

Background:

• Accurate pain assessment is crucial for effective pain management in clinical settings.
• Objective pain detection from physiological data aids caregivers, especially for non-self-reporting patients.
• Galvanic skin response (GSR) measures sweat gland activity, reflecting emotional states and anxiety linked to pain levels.

Purpose of the Study:

• To develop an automatic pain assessment tool utilizing GSR signals.
• To predict varying pain intensities in noncommunicative, postoperative adult patients.
• To establish the efficacy of GSR-based pain detection in a real-world clinical population.

Main Methods:

• Collected GSR data from 25 postoperative patients (aged 23-89) reporting moderate to high pain.
• Utilized Empatica E4 wristband for data acquisition during low-intensity activities.
• Employed machine learning algorithms (Random Forest, k-Nearest Neighbor) and feature selection (mean decrease impurity) to build predictive models.
• Correlated GSR data with patient self-reports (Numeric Rating Scale) and compared results to prior studies.

Main Results:

• Developed binary classification models to distinguish baseline from different pain levels (PL1-PL4).
• Achieved high prediction accuracies: Random Forest yielded 86.0% (BL vs PL1), 70.0% (BL vs PL2), and 61.5% (BL vs PL4).
• K-Nearest Neighbor achieved 72.1% accuracy for BL vs PL3, outperforming previous research on real patient data.

Conclusions:

• This is the first study to propose and validate a GSR-based pain assessment tool for postoperative adult patients.
• Feature selection algorithms successfully identified key features for pain intensity prediction.
• The findings demonstrate the potential of GSR signals for objective, automated pain assessment in clinical practice.