Network physiology has increasingly revealed the dynamic interactions between physiological systems, with heart–brain coupling receiving particular attention. In this study, we propose the amplitude-aware symbolic transfer entropy (AASTE) algorithm to enhance the analysis of heart–brain coupling dynamics, specifically focusing on EEG and ECG interactions across different sleep stages: wakefulness (W), light sleep (N1), intermediate sleep (N2), and deep sleep (N3). Our results show that the transfer entropy from ECG to EEG is consistently higher than from EEG to ECG, suggesting a stronger regulatory influence of heart signals on brain activity, particularly during wakefulness. Furthermore, sleep depth significantly impacts heart–brain information coupling, with interactions gradually weakening as sleep deepens, while the central brain regions (C3, C4) play a crucial role in reflecting these changes. This study explores heart–brain interactions across sleep stages and demonstrates AASTE’s potential for quantifying directional information flow in physiological systems, providing a comprehensive tool for investigating the dynamic transmission of physiological information.
Heart–brain coupling, regulated by the autonomic and central nervous systems, plays a vital role in maintaining physiological homeostasis, emotional regulation, and cognitive function. Understanding this dynamic interaction offers valuable insights into the coordination of physiological processes and the diagnosis of cardiovascular and neurological disorders. In this study, we applied symbolic phase transfer entropy to analyze directional information flow between the heart and brain during different sleep stages. The results revealed that heart-to-brain information flow consistently dominates across all frequency bands and sleep stages, with stronger coupling in lower-frequency bands during sleep and diminished coupling in higher-frequency bands associated with wakefulness. Significant differences in transfer entropy were observed between wakefulness and sleep, particularly in the δ and β bands, with consistent trends across electrodes C3, C4, O1, and O2. Using the identified transfer entropy differences as features, a K-nearest neighbors classifier achieved a mean accuracy of 85.43% and an AUC of 0.9276, demonstrating robust performance in distinguishing wakefulness from sleep states. These findings highlight the innovative application of directional information flow metrics in understanding heart–brain interactions and their potential for advancing sleep state diagnostics.
Depression is a severe mental disorder, and patients suffering from depression differ significantly from those in the control group in terms of electroencephalogram (EEG) signal complexity. Although most of the existing studies have focused on overall complexity analysis, very few have explored the complexity characteristics from a decomposition perspective. In this paper, we propose to apply the fast Fourier transform to the decomposition method to resolve the nonlinear feature differences (total differences, TDs) between task and resting eye-open states and decompose them into spectral terms (STs), phase terms (PTs), and spectral interaction terms (SITs). The selected nonlinear features include Lempel–Ziv complexity (LZC), permutation entropy (PE), and basic scale entropy (BSE). The experimental data were obtained from public datasets on the OpenNeuro website, including the depression resting-state and task-state datasets, involving EEG data from 46 depressed patients and 74 controls. The results indicated that TDs and STs were significantly lower in the depressed group than in the control group in the frontal region, while the SIT was significantly higher in the frontal region and lower in the central region. The PT in the depressed group was lower in the frontal region but higher in the central and temporal regions. In addition, the two groups exhibited opposite trends in the SIT across the two states. Entropy decomposition of the LZC, PE, and BSE differences effectively differentiated depressed patients, with BSE differences distinguishing the highest number of channels. These results may provide an important reference for the clinical diagnosis and treatment of depression.
6 April 2026 in IEEE Transactions on Neural Systems and Rehabilitation Engineering
Clarifying the mechanisms of altered brain network communication in Parkinson’s disease is crucial, yet conventional connectivity measures do not distinguish whether directed information flow is driven by phase synchrony or spectral amplitude modulation. To address this, we introduce Decomposed Transfer Entropy (DTE), which represents directed information flow as a mechanistic profile comprising phase, spectral, and interaction components. Simulations indicate that DTE identifies distinct coupling types and remains robust under noise; we further benchmark DTE against representative baseline measures and assess parameter sensitivity. Applied to θ band resting EEG, DTE reveals that the unmedicated (PD-OFF) state is characterized by a phase-dominant reweighting originating from the frontal midline hub Fz; medication shifts this mechanistic composition toward that of healthy reference. Hub-centric composition metrics derived from DTE discriminate PD-OFF from healthy controls, and the normalization of phase dominance relates to motor improvement (ΔUPDRS–III). Overall, DTE offers a mechanism-aware perspective on PD pathophysiology, suggesting that therapeutic benefit aligns with normalization of mechanistic composition rather than simple changes in connection strength, and provides a promising route toward mechanism-oriented biomarkers.