Objective, To address the limitation that existing brain network analysis methods relying on relatively simplistic data processing models fail to capture the multimodal information embedded in neural signals, this study introduces a multidimensional information transfer measurement method - Kalman Time Phase Transfer Entropy (nKTE-TP), and applies it to Enhanced Brain Computer Interface (BCI) decoding. Approach, Built upon the transfer entropy algorithm, this method leverages the Kalman filter to dynamically weight signal amplitude and phase information, while taking both the temporal and spectral characteristics of signals into account, thus effectively integrating amplitude energy-driven features with phase temporal properties. The effectiveness of the nKTE-TP method was validated through multivariable autoregressive model (MVAR) simulation experiments and three different types of publicly available BCI datasets, and compared with existing algorithms. Main results, MVAR simulation results demonstrate that this method robustly assesses causal information and exhibits superior stability and accuracy in evaluating complex systems compared to existing algorithms. In addition, the validation results of three publicly available BCI datasets show that compared with other causal algorithms, the nKTE-TP network features exhibit higher discriminative ability and generate more robust recognition models. Significance, This method offers a more comprehensive metric for investigating information dependence and integration across different brain regions, while simultaneously providing an effective decoding strategy for BCI neural signal analysis.
Zhang et al. (Thu,) studied this question.