Related Experiment Video
Updated: Aug 15, 2025
09:16
Multi-Modal Signals for Analyzing Pain Responses to Thermal and Electrical Stimuli
Published on: April 5, 2019
10.8K
External validation of binary machine learning models for pain intensity perception classification from EEG in
Tyler Mari1, Oda Asgard2, Jessica Henderson2
1Department of Psychology, Institute of Population Health, University of Liverpool, 2.21 Eleanor Rathbone Building, Bedford Street South, Liverpool, L69 7ZA, UK. Tyler.Mari@liverpool.ac.uk.
Scientific Reports
|January 5, 2023
Summary
This study externally validated machine learning (ML) and electroencephalography (EEG) for classifying pain intensity. ML models showed promising generalization to new data, demonstrating ML
Area of Science:
- Neuroscience
- Computational Neuroscience
- Biomedical Engineering
Background:
- Machine learning (ML) and electroencephalography (EEG) show promise for objective pain intensity assessment, particularly when self-report is not feasible.
- Current research in this field is constrained by a lack of external validation, limiting the generalizability of findings to novel datasets and experimental paradigms.
Purpose of the Study:
- To conduct the first external validation study for classifying pain intensity using EEG and ML.
- To assess the performance of ML models trained on EEG data in classifying high versus low pain intensities in independent datasets with varying stimulation parameters.
Main Methods:
- Two independent EEG experiments were conducted with healthy participants subjected to high and low intensity pneumatic pressure stimuli.
- Time-frequency features from peri-stimulus EEG data were extracted on a single-trial basis. Feature selection identified relevant features from frontal, central, and parietal regions.
- Machine learning models, including Random Forest, were trained and validated on distinct datasets, including novel stimulation parameters for external validation.
Main Results:
- Machine learning models significantly outperformed chance in classifying pain intensity.
- The Random Forest model achieved the highest accuracies: 73.18% for cross-validation, 68.32% for external validation one, and 60.42% for external validation two.
- Performance demonstrated generalization to novel samples and experimental paradigms, indicating robustness of the ML-EEG approach.
Conclusions:
- This study provides the first rigorous external validation of ML and EEG for pain intensity classification.
- The findings demonstrate the clinical potential of ML-EEG for objective pain assessment, showing promising and generalizable performance.
- This research offers reliable estimates of ML's capability in pain classification, paving the way for more robust clinical applications.
