Frontiers in network physiology最新文献

Understanding the underlying dynamics of complex real-world systems, such as neurophysiological and climate systems, requires quantifying the functional interactions between the system units under different scenarios. This tutorial paper offers a comprehensive description to time, frequency and information-theoretic domain measures for assessing the interdependence between pairs of time series describing the dynamical activities of physical systems, supporting flexible and robust analyses of statistical dependencies and directional relationships. Classical time and frequency domain correlation-based measures, as well as directional approaches derived from the notion of Granger causality, are introduced and discussed, along with information-theoretic measures of symmetrical and directional coupling. Both linear model-based and non-linear model-free estimation approaches are thoroughly described, the latter including binning, permutation, and nearest-neighbour estimators. Special emphasis is placed on the description of a unified framework that establishes a connection between causal and symmetric, as well as spectral and information-theoretic measures. This framework enables the frequency-specific representation of information-theoretic metrics, allowing for a detailed investigation of oscillatory components in bivariate systems. The practical computation of the interaction measures is favoured by presenting a software toolbox and two exemplary applications to cardiovascular and climate data. By bridging theoretical concepts with practical tools, this work enables researchers to effectively investigate a wide range of dynamical behaviours in various real-world scenarios in Network Physiology and beyond.

Artificial Intelligence and machine learning tools have brought a revolution in the healthcare sector. This has allowed healthcare providers, patients, and public to be at pole position -amidst the key consideration and barriers-to attain precision and personalized medicine. Deep Learning (DL) is a branch of machine learning and AI that has become transformative for healthcare and biomedicine, providing the ability to analyze large, complicated data, capture abstract patterns, and present fast and accurate predictions. DL models are based on complex neural networks that emulate biological neural networks. In this paper, our goal is to evaluate DL algorithms in clinical trials stratified per cancer type and present future perspectives on the most promising DL approaches. We systematically reviewed articles on deep learning in cancer diagnostics in studies published in the Pubmed database. The searched literature included two types of articles, clinical trials, and randomized controlled trials. The deep learning algorithms used in the targeted literature are reviewed, and then we evaluated the performance of the algorithms used in disease prediction and prognosis. We aim to highlight the promising DL approaches reported per cancer type. Finally, we present current limitations and potential recommendations in large-scale implementation of deep learning and AI in cancer care.

Introduction: Cardiovascular diseases, particularly Coronary Artery Disease (CAD), remain a leading cause of mortality worldwide. Invasive angiography, while accurate, is costly and risky. This study proposes a non-invasive, interpretable CAD prediction framework using the Z-Alizadeh Sani dataset.

Methods: A hybrid decision tree-AdaBoost method is employed to select 30 clinically relevant features. To prevent data leakage, SMOTE oversampling is applied exclusively within each training fold of a 10-fold cross-validation pipeline. The Support Vector Machine (SVM) model is optimized using Bayesian hyperparameter tuning and compared against Sea Lion Optimization Algorithm (SLOA) and grid search. SHapley Additive exPlanations (SHAP) analysis is utilized to interpret the feature contributions.

Results: The SVM_Bayesian model achieves 97.67% accuracy, 95.45% precision, 100.00% sensitivity, 97.67% F1-score, and 99.00% AUC, outperforming logistic regression (93.02% accuracy, 92.68% F1-score), random forest (95.45% accuracy, 93.33% F1-score), standard SVM (77.00% accuracy), and SLOA-optimized SVM (93.02% accuracy). Ablation studies and Wilcoxon signed-rank tests confirm the statistical superiority of the proposed model.

Discussion: SHAP analysis reveals clinically meaningful feature contributions (e.g., Typical Chest Pain, Age, EFTTE). 95% bootstrap confidence intervals and temporal generalization on an independent test set ensure robustness and prevent overfitting. Future work includes validation on external real-world datasets. This framework provides a transparent, generalizable, and clinically actionable tool for CAD risk stratification, aligned with the principles of network physiology by focusing on interconnected cardiovascular features in predicting systemic disease.

The study of neuron models and their networks is a riveting topic for many researchers worldwide because it allows to glimpse the fundamental processes using accessible methodology. The paper considers dynamics of small networks of Hodkin-Huxley neurons, namely a chain of three neurons and a small-world-like network of seven neurons. The ensembles of neurons are represented by systems of ordinary differential equations, so the research has been conducted numerically. It has been found that complex quasi-periodic and chaotic regimes may arise in the systems, and the existense of such regimes is caused by the inner parameters of the systems, such as individual currents of the neurons and the coupling between them. This research contributes to the fundamental understanding of signal propagation in networks of neuron models and may provide insight into the physiology of real neuronal systems.

Background: Severe mitral valve regurgitation requires comprehensive evaluation for optimal treatment. Initial screening uses transthoracic echocardiography (TTE), followed by transesophageal echocardiography (TEE) to determine eligibility for adequate intervention. Mitral Transcatheter Edge-to-Edge Repair (M-TEER) indications are based on detailed and quality valve and sub-valvular apparatus assessment, including anatomy and regurgitation pathophysiology.

Aim: To develop AI algorithms for standardizing M-TEER eligibility assessment using TTE and TEE echocardiograms, supporting all stages of mitral valve regurgitation evaluation to assist non-expert centers throughout the entire process, from severe mitral valve regurgitation diagnostic to M-TEER procedure.

Methods: Three deep learning algorithms were developed using echocardiographic data from M-TEER patients performed at Montreal Heart Institute (2018-2025). 1. ECHO-PREP was trained to identify key diagnostic views in TTE (n = 530) and diagnostic and procedural views in TEE (n = 2,222) examinations to determine the level of quality images needed to do a M-TEER. 2. 4D TEE segmentation with automated mitral valve area (MVA) quantification (n = 221), and 3. 2D TEE scallop-level segmentation of leaflets and sub-valvular structures (n = 992).

Results: Preliminary results on test sets showed 95.7% accuracy in TTE view classification and 91% accuracy for TEE view classification. The 4D segmentation module demonstrated excellent agreement with manual MVA measurements (R = 0.84, p < 0.001), successfully discriminating patients undergoing M-TEER from those referred for surgical replacement (p = 0.046 for AI predictions). The 2D scallop-level analysis achieved a mean Dice score of 0.534 across 11 anatomical structures, with better performance in commonly represented configurations (e.g., A2-P2, P1-A2-P3).

Conclusion: ECHO-PREP demonstrates the feasibility of an integrated AI-assisted workflow for MR assessment, combining quality control, dynamic 4D valve quantification, and scallop-level anatomy interpretation. These results support the potential of AI to standardize M-TEER eligibility, reduce inter-observer variability, and provide decision support across centers with different levels of expertise.

Sudden cardiac death (SCD) is often precipitated by reentrant arrhythmias such as ventricular tachycardia (VT) and ventricular fibrillation (VF), whose underlying dynamics are frequently sustained by spiral waves of electrical activity. Disrupting these waves can restore normal rhythm, but conventional low-energy pacing strategies are often ineffective in VF, where high-frequency, multi-wave interactions dominate. Resonant feedback-controlled antitachycardia pacing (rF-ATP), which times global electrical stimuli based on real-time feedback from the tissue, has been shown to robustly terminate single spirals under diverse conditions. However, its impact on interacting spiral waves-arguably a more realistic substrate for life-threatening arrhythmias-remains unexplored. Here, we use numerical simulations to investigate the effect of rF-ATP on figure-of-eight reentry, a clinically relevant configuration consisting of two counter-rotating spirals. We show that rF-ATP consistently terminates this pattern, regardless of feedback point location, through two distinct dynamical pathways: mutual collision of phase singularities or annihilation at inexcitable boundaries. We further demonstrate the method's efficacy across variations in feedback point and spiral arrangement, indicating robustness to geometrical and positional heterogeneity. These results highlight rF-ATP as a promising low-energy intervention for complex reentrant structures and provide mechanistic insight into feedback-driven control of multi-core spiral wave dynamics in cardiac tissue.

Cardiac dynamics is governed by complex electrical wave patterns, with disruptions leading to pathological conditions like atrial or ventricular fibrillation. Experimentally electrical excitation waves can be made visible by optical mapping using fluorescent dyes. While this imaging technique has enabled detailed studies of cardiac wave dynamics, the manual analysis of activation and phase maps often limits the ability to systematically identify and quantify wave patterns. This study employs a wave tracking algorithm that constructs a graph-based representation of wave dynamics. With that the algorithm detects key events such as wave emergence, splitting, and merging. Applied to both simulated cardiac tissue and experimental data from cell cultures, the algorithm identifies and quantifies wave patterns as wave event networks. Initial results demonstrate its utility in filtering for and focusing on dominant dynamics, providing a robust tool for analyzing cardiac wave patterns. This approach offers potential applications, e.g., to study the effects of external stimuli on cardiac excitation patterns and to better understand the mechanisms involved.

It has been shown before that two species of diffusing particles can separate from each other by the mechanism of reciprocally concentration-dependent diffusivity: the presence of one species amplifies the diffusion coefficient of the respective other one, causing the two densities of particles to separate spontaneously. In a minimal model, this could be observed with a quadratic dependence of the diffusion coefficient on the density of the other species. Here, we consider a more realistic sigmoidal dependence as a logistic function on the other particle's density averaged over a finite sensing radius. The sigmoidal dependence accounts for the saturation effects of the diffusion coefficients, which cannot grow without bounds. We show that sigmoidal (logistic) cross-diffusion leads to a new regime in which a homogeneous disordered (well-mixed) state and a spontaneously separated ordered (demixed) state coexist, forming two long-lived metastable configurations. In systems with a finite number of particles, random fluctuations induce repeated transitions between these two states. By tracking an order parameter that distinguishes mixed from demixed phases, we measure the corresponding mean residence in each state and demonstrate that one lifetime increases and the other decreases as the logistic coupling parameter is varied. The system thus displays typical features of a first-order phase transition, including hysteresis for large particle numbers. In addition, we compute the correlation time of the order parameter and show that it exhibits a pronounced maximum within the bistable parameter range, growing exponentially with the total particle number.

In this perspective, we introduce the Precision Principle as a unifying theoretical framework to explain self-organization across biological systems. Drawing from neurobiology, systems theory, and computational modeling, we propose that precision, understood as constraint-driven coherence, is the key force shaping the architecture, function, and evolution of nervous systems. We identify three interrelated domains: Structural Precision (efficient, modular wiring), Functional Precision (adaptive, context-sensitive circuit deployment), and Evolutionary Precision (selection-guided architectural refinement). Each domain is grounded in local operations such as spatial and temporal averaging, multiplicative co-activation, and threshold gating, which enable biological systems to achieve robust organization without centralized control. Within this framework, we introduce the Precision Coefficient, $P\left(z\right)=C\left(z\right)-\alpha R\left(z\right)$ , which formalizes the balance between network coherence and resource cost and serves as a simple quantitative outline of the principle. Conceptually, this formalism aligns with established learning mechanisms: Hebbian reinforcement provides the local substrate for weight changes, while winner-take-all and k-winners competition selectively eliminates weaker synapses, together increasing $C\left(z\right)$ and reducing redundancy within $R\left(z\right)$ . Rather than framing the theory in opposition to existing models, we aim to establish the Precision Principle as an original, integrative lens for understanding how systems sustain efficiency, flexibility, and resilience. We hope the framework inspires new research into neural plasticity, development, and artificial systems, by centering internal coherence, not prediction or control, as the primary driver of self-organizing intelligence.

Aging is associated with a decline in inter-muscular coordination and overall functional capacity. While the benefits of exercise on individual physiological systems are well established, it remains unclear whether regular training can also enhance inter-muscular network interactions and counteract age-related decline. Using a Network Physiology approach, this Case Report investigates the effects of a home-based exercise program on inter-muscular coordination in two older adults. Two older adults (aged 69 and 73) completed a 12-week program that included twice-weekly virtual group sessions, and one weekly session of moderate-intensity aerobic exercise (30 min). Before and after the intervention, participants underwent a maximal cardiopulmonary exercise test (CPET) on a motorized treadmill. During the CPET, surface electromyography (EMG) was recorded from the left and right rectus femoris and biceps femoris. Inter-muscular coordination was quantified using the Amplitude-Amplitude Cross-Frequency Coupling (ACFC) method. Ten time series of EMG band power were extracted for each muscle, representing distinct neuromuscular processes. Pearson's cross-correlation was then computed for each pair of EMG band power time series across all muscles. Pre-Intervention, both participants showed low overall link strength across all sub-networks. Post-Intervention, there was a pronounced (∼400%) increase in average link strength across all sub-networks in both participants, primarily reflecting enhanced synchronization between distinct frequency bands across the rectus femoris and biceps femoris. These preliminary findings suggest that structured exercise may enhance inter-muscular network coordination in older adults. ACFC-derived network measures offer a promising tool for detecting early age-related decline and evaluating neuromuscular adaptations to exercise interventions.