Answers are grounded in this paper and cite their sources.
Classification model achieved 93.41% accuracy in EEG pain detection, highlighting improved feature selection and temporal resolution.
Introduction: Objective assessment of pain is important for clinical diagnosis and therapeutic intervention. But traditional scales like the Visual Analogue Scale (VAS) or Numeric Rating Scale (NRS) depend on patients reporting their own symptoms, which may be difficult for some people. Methods: An EEG short-time window pain detection model with Fusion of Mutual Information‐Based Recursive Feature Elimination (MI-RFE) + Gradient Boosting (GB) is proposed in this study. By comparing slide window lengths from one to five seconds, we found that a one-second window provided the optimal balance between capturing transient pain signals and increasing the size of the dataset. The dataset is enlarged 15 times by a one-second window. This paper proposes a classification model that fuses the Mutual Information-based RFE method (MI-RFE) with the Gradient Boosting classifier. The MI-RFE is used to select high-dimensional features, then the resulting subset is used to train a Gradient Boosting classifier. Results: Finally, this EEG short-time window pain detection model of MI-RFE + GB achieved an accuracy of 93.41%, an F1 score of approximately 93.54%, and a maximum ROC AUC of 97.96%, better than other methods such as SVM. Discussion: The results show that the one-second window more effectively captures transient pain signals while maintaining temporal resolution and improving model training accuracy. MI-RFE + Gradient Boosting model fully exploits the non-linear relationship between EEG features and pain labels, demonstrating strong generalization capabilities. Conclusion: This study provides a novel technical solution with significant patent potential, offering practical insights for developing objective pain assessment systems and demonstrating promising applications in clinical pain management.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: