Recurrence Quantification Analysis of EEG signals : Parameters Selection and Correlations with Spectral Features

Recurrence Quantification Analysis (RQA) offers nonlinear features to characterize the temporal dynamics of complex systems. While increasingly applied to EEG signal analysis, its use in this context remains challenged by the lack of standardized parameter selection procedures, compromising reproducibility and comparability across studies. This work provides a systematic evaluation of RQA parameter configurations, including embedding dimension, time delay, recurrence radius, and minimal diagonal line length, applied to resting-state EEG under eyes-open and eyes-closed conditions. An adaptive, datadriven strategy is proposed to optimize parameter selection, notably with no embedding and using an approach derived from kernel density estimation for recurrence radius adjustment. Statistical analyses demonstrate the robustness of the determinism (DET) measure in detecting condition-related temporal structure differences. Complementary principal component analyses (PCA) confirm that RQA features capture information distinct from classical spectral features, underscoring the relevance of combining nonlinear and spectral analyses. The proposed framework enhances methodological rigor, reproducibility, and applicability of RQA in EEG studies.