Related Experiment Video
Updated: Jul 16, 2026
09:16
Multi-Modal Signals for Analyzing Pain Responses to Thermal and Electrical Stimuli
Published on: April 5, 2019
Physiological Data Analysis Framework for Pain Prediction in Physical Rehabilitation
Abdel Hiram Cital Duarte1, Gilberto Borrego2, Samuel González-López3
1Department of Electrical and Electronics, Instituto Tecnológico de Sonora, Ciudad Obregón 85130, Sonora, Mexico.
Sensors (Basel, Switzerland)
|July 15, 2026
Summary
Machine learning models using heart rate (HR) and other non-invasive sensors can detect pain during physical rehabilitation. This approach offers reliable, low-cost pain monitoring for personalized therapy, especially in telerehabilitation.
Area of Science:
- Biomedical Engineering
- Rehabilitation Science
- Machine Learning
Background:
- Pain prediction in physical rehabilitation is difficult due to subjective reporting and patient variability.
- Telerehabilitation settings exacerbate challenges in pain assessment, particularly when self-reports are unreliable.
Purpose of the Study:
- To assess if machine learning models using heart rate (HR), heart rate variability (HRV), and peripheral oxygen saturation (SpO2) can reliably detect pain during rehabilitation.
- To investigate the utility of low-cost, non-invasive physiological markers for pain detection without self-reporting, applicable to both clinical and home settings.
Main Methods:
- Collected HR, HRV, and SpO2 data from 25 participants undergoing lower-limb rehabilitation.
- Employed machine learning models including linear regression (LR), random forest (RF), and artificial neural networks (ANNs).
- Processed physiological signals using temporal filtering, quality screening, and imputation strategies before model training and evaluation.
Main Results:
- Linear regression showed limited pain prediction capability.
- Random Forest achieved high accuracy (97.77%) in detecting low-pain episodes.
- ANN models provided a more balanced three-class pain profile but were sensitive to data imbalance.
Conclusions:
- Physiological markers combined with machine learning show promise for detecting pain during real rehabilitation sessions.
- This non-invasive, sensor-based approach can facilitate personalized rehabilitation and proactive therapy adjustments.
- Findings are applicable to both in-clinic and telerehabilitation, enhancing patient adherence and outcomes.
