Modified Feature Selection for Improved Classification of Resting-State Raw EEG Signals in Chronic Knee Pain

TL;DR

This work tackles the absence of objective pain biomarkers by predicting chronic knee pain from resting-state EEG using a compact connectivity signature. It introduces mSFFS, a modified feature-selection algorithm that explores multiple search paths to avoid local minima, and demonstrates superior class separability and classification accuracy ($0.975$ test, $0.960$ CV) with a 20-feature connectivity set derived from ciPLV across five frequency bands. Through t-SNE visualization and Bhattacharyya distance analysis, the study shows clearer class separation and greater separability than SFFS, SHAP-based ranking, and swarm-based methods. The results suggest a practical, low-cost EEG-based diagnostic approach and offer neurophysiological insights into pain-related brain connectivity, with plans to extend to larger, cross-source datasets and domain-adaptation techniques.

Abstract

\textit{Objective:} Diagnosing pain in research and clinical practices still relies on self-report. This study aims to develop an automatic approach that works on resting-state raw EEG data for chronic knee pain prediction. \textit{Method:} A new feature selection algorithm called ``modified Sequential Floating Forward Selection'' (mSFFS) is proposed. The improved feature selection scheme can better avoid local minima and explore alternative search routes. \textit{Results:} The feature selection obtained by mSFFS displays better class separability as indicated by the Bhattacharyya distance measures and better visualization results. It also outperforms selections generated by other benchmark methods, boosting the test accuracy to 97.5\%. \textit{Conclusion:} The improved feature selection searches out a compact, effective subset of connectivity features that produces competitive performance on chronic knee pain prediction. \textit{Significance:} We have shown that an automatic approach can be employed to find a compact connectivity feature set that effectively predicts chronic knee pain from EEG. It may shed light on the research of chronic pains and lead to future clinical solutions for diagnosis and treatment.

Figures (6)

• Figure 1: Classification accuracy scores when using an increasing number of mSFFS selected features. The shadow area gives the confidence intervals of the cross-validation scores.
• Figure 2: Testing scores using an increasing number of features. The shadow areas represent the confidence intervals for 10-fold tests.
• Figure 3: Incremental differences of feature selections by SFFS and mSFFS.
• Figure 4: Jaccard index of the two evolving feature selections over time.
• Figure 5: Mean absolute SHAP values of the features selected by three feature selection schemes: (a) top-20 SHAP; (b) SFFS, 20 features; (c) SFFS, 20 features.