Cross-Frequency Bispectral EEG Analysis of Reach-to-Grasp Planning and Execution

TL;DR

This work extends higher-order spectral analysis to ecologically valid motor tasks by applying cross-frequency bispectrum to EEG during reach-to-grasp, differentiating planning from execution and decoding grip type. Using 25 band-pair interactions and both magnitude- and phase-based features, the authors demonstrate that execution engenders stronger, more focal nonlinear coupling, particularly in $eta$ and $\\gamma$ bands, with prefrontal and occipital regions implicated in the network. The approach achieves robust within-subject classification and generalizes to unseen data, especially when using permutation-derived top features, and reveals a progression from distributed planning dynamics to high-frequency, task-specific coupling during execution. These findings advance the interpretability and practicality of bispectral EEG markers for brain–computer interfaces and neuroprosthetic control, highlighting specific band-pair interactions and spatial patterns as targets for future decoding algorithms. Overall, the study provides a methodological and conceptual framework for incorporating cross-frequency nonlinear dynamics into neural decoding of naturalistic grasping actions.

Abstract

Neural control of grasping arises from nonlinear interactions across multiple brain rhythms, yet EEG-based motor decoding has largely relied on linear, second-order spectral features. Here, we examine whether higher-order cross-frequency dynamics distinguish motor planning from execution during natural reach-to-grasp behavior. EEG was recorded in a cue-based paradigm during executed precision and power grips, enabling stage-resolved analysis of preparatory and execution-related neural activity. Cross-frequency bispectral analysis was used to compute bicoherence matrices across canonical frequency band pairs, from which magnitude- and phase-based features were extracted. Classification, permutation-based feature selection, and within-subject statistical testing showed that execution is characterized by substantially stronger and more discriminative nonlinear coupling than planning, with dominant contributions from beta- and gamma-driven interactions. In contrast, decoding of precision versus power grips achieved comparable performance during planning and execution, indicating that grasp-type representations emerge during planning and persist into execution. Spatial and spectral analyses further revealed that informative bispectral features reflect coordinated activity across prefrontal, central, and occipital regions. Despite substantial feature redundancy, effective dimensionality reduction preserved decoding performance. Together, these findings demonstrate that nonlinear cross-frequency coupling provides an interpretable and robust marker of motor planning and execution, extending bispectral EEG analysis to ecologically valid grasping and supporting its relevance for brain-computer interfaces and neuroprosthetic control.

Figures (6)

• Figure 1: Experimental Setup and Data Collection Protocol for Vision-Based Grasping Platform. (a) Depicts a participant wearing the EEG recording system during the experimental protocol. The top view of the turntable rotates to present one of three conditions at random: bottle, pen, or empty. (b) Single trial structure and outline of EEG recording session, beginning with the planning stage then the reach-to-grasp execution following audio cue onset.
• Figure 2: Overview of the cross-frequency bispectral analysis and classification pipeline. (a) EEG preprocessing and bispectrum computation. (b) Derivation of cross-frequency driver–band matrices and extraction of phase- and magnitude-based bispectral features across all channels and band-pairs. (c) Feature selection and multi-task classification.
• Figure 3: Subject-averaged feature maps across channels and frequency band pairs. Each large panel (A-I) corresponds to one feature: A: mean, B: max, C: sum, D: entropy, E: sine of the biphase, F: cosine of the biphase, G: phase-coupling concentration $R$, H: circular variance, and I: phase entropy. Within each feature block, six subplots are for each task condition: NM, bottle, and pen in planning (a, c, e), and in execution (b, d, f).
• Figure 4: Classification accuracies (10-fold cross-validation) across frequency bands ($\delta$, $\theta$, $\alpha$, $\beta$, $\gamma$) for Planning (blue) and execution (orange) stages in four task settings: (a) Power grasp vs. No-execution, (b) Precision grasp vs. No-execution, (c) Power vs. Precision grasp, and (d) Multiclass (Power, Precision, No-execution). Each boxplot summarizes subject-level accuracies ($N=10$). Horizontal bars above each band indicate paired comparisons between stages, with statistical outcomes corrected for multiple comparisons (FDR): ns = non-significant, and ** is significant $p < 0.05$. Effect sizes (Cohen’s $d$) are shown in parentheses.
• Figure 5: Most frequently selected feature components across all subjects, conditions, and task phases. (a) Top 10 EEG channels, (b) Top 10 band-pair interactions, and (c) Top bispectral feature types, based on their global frequency of selection across all classification models. Bar heights indicate the total number of times each component appeared among the top-ranked features. This global summary highlights the most consistently informative spatial (channel), spectral (band-pair), and statistical (feature type) domains in the bispectral feature space across subjects.