Abstract:Physiological responses to pain have received increasing attention among researchers for developing an automated pain recognition sensing system. Though less explored, Blood Volume Pulse (BVP) is one of the candidate physiological measures that could help objective pain assessment. In this study, we applied machine learning techniques on BVP signals to device a non-invasive modality for pain sensing. Thirty-two healthy subjects participated in this study. First, we investigated a novel set of time-domain, frequency-domain and nonlinear dynamics features that could potentially be sensitive to pain. These include 24 features from BVP signals and 20 additional features from Inter-beat Intervals (IBIs) derived from the same BVP signals. Utilizing these features, we built machine learning models for detecting the presence of pain and its intensity. We explored different machine learning models, including Logistic Regression, Random Forest, Support Vector Machines, Adaptive Boosting (AdaBoost) and Extreme Gradient Boosting (XGBoost). Among them, we found that the XGBoost offered the best model performance for both pain classification and pain intensity estimation tasks. The ROC-AUC of the XGBoost model to detect low pain, medium pain and high pain with no pain as the baseline were 80.06 %, 85.81 %, and 90.05 % respectively. Moreover, the XGboost classifier distinguished medium pain from high pain with ROC-AUC of 91%. For the multi-class classification among three pain levels, the XGBoost offered the best performance with an average F1-score of 80.03%. Our results suggest that BVP signal together with machine learning algorithms is a promising physiological measurement for automated pain assessment. This work will have a national impact on accurate pain assessment, effective pain management, reducing drug-seeking behavior among patients, and addressing national opioid crisis.
What problem does this paper attempt to address?
The problem this paper attempts to address is the automatic recognition of the presence and intensity of pain from Blood Volume Pulse (BVP) signals using machine learning techniques. Specifically, the researchers aim to develop a non-invasive method to objectively assess pain using BVP signals and machine learning algorithms, thereby improving the accuracy and effectiveness of pain management.
### Main Research Objectives:
1. **Explore the potential of BVP signals in pain assessment**: The researchers aim to analyze BVP signals to extract a series of time-domain, frequency-domain, and nonlinear dynamic features that may be sensitive to pain.
2. **Build machine learning models for pain detection and intensity estimation**: Using the extracted features, train various machine learning models (such as logistic regression, random forest, support vector machine, adaptive boosting, and extreme gradient boosting) to detect the presence of pain and estimate its intensity.
3. **Evaluate the performance of different models**: Compare the performance of different models through cross-validation and various evaluation metrics (such as ROC-AUC, F1 score, etc.) to identify the best model.
### Research Background:
- **Importance of pain assessment**: Pain is a major symptom of many medical conditions, and effective pain management requires accurate pain assessment. However, current methods mainly rely on patient self-reporting, which has limitations, especially in certain patient groups (such as those with dementia or paralysis).
- **Application of physiological signals**: The autonomic nervous system plays an important role in the pain experience, and these changes can be measured through physiological signals (such as EEG, ECG, BVP, and skin conductance) to achieve automated pain assessment.
- **Limitations of existing research**: Although there have been studies using other physiological signals (such as skin conductance) for pain assessment, the application of BVP signals in this area has been less explored.
### Research Methods:
- **Data collection**: Collect BVP signals from 32 healthy subjects through a cold pressor test, recording signals in baseline and pain states.
- **Signal preprocessing**: Filter and denoise the BVP signals, and extract Inter-beat Intervals (IBIs).
- **Feature extraction**: Extract 44 features from BVP signals and IBIs, including time-domain, frequency-domain, and nonlinear dynamic features.
- **Model training and evaluation**: Use various machine learning models for binary and multi-class classification tasks, evaluating the models' performance in pain detection and intensity estimation.
### Main Findings:
- **XGBoost model performs best**: In pain detection and intensity estimation tasks, the XGBoost model performed the best, with ROC-AUCs of 80.06%, 85.81%, and 90.05% (corresponding to low pain, medium pain, and high pain, respectively).
- **Multi-class tasks**: The XGBoost model also performed well in the three-level pain classification task, with an average F1 score of 80.03% and a ROC-AUC of 92.25%.
### Significance:
- **Clinical application prospects**: The study results indicate that BVP signals combined with machine learning algorithms are a promising physiological measurement method for automated pain assessment, which can help improve the accuracy and effectiveness of pain management.
- **Social impact**: This method is expected to reduce drug abuse, address the national opioid crisis, and improve patients' quality of life.