Editorial: New theories, models, and AI methods of brain dynamics, brain decoding and neuromodulation

Yuzhu Guo, Yang Li, Hua‐Liang Wei, Yifan ZhaoFrontiers in Neuroscience 2023 · DOI · OpenAlex

Introduction: The Challenge of the Dynamic Brain

The human brain is a pinnacle of biological complexity, functioning as a highly dynamic system that integrates signals across diverse spatial and temporal scales. Understanding these dynamics is not merely an academic exercise; it is a prerequisite for advancing neuromodulation—the use of external stimuli to alter neural activity for therapeutic purposes. However, researchers face significant hurdles. Brain signals are inherently nonlinear and non-stationary, containing a wealth of information embedded in temporal, spatial, and frequency domains. Furthermore, while modern AI and deep learning models have shown great promise in decoding these signals, they often operate as "black boxes," lacking the physiological interpretability required for clinical safety and the optimization of stimulation parameters.

This Research Topic explores the intersection of medicine, neurobiology, and engineering to bridge this gap. By combining top-down and bottom-up methodologies, the featured studies aim to produce robust, interpretable results that reveal the underlying mechanisms of brain dynamics, with a specific focus on brain decoding and neuromodulation.

Methodological Diversity in Brain Decoding

The collection highlights four distinct approaches to capturing and influencing brain dynamics, demonstrating the interdisciplinary nature of the field:

1. Cross-Domain Data Augmentation

In many clinical scenarios, high-quality neural data is scarce. Liu et al. addressed this by developing a cross-domain data augmentation method. By synthesizing richer datasets that combine spatial and temporal features, they improved the identification of brain disorders. This approach suggests that even with limited raw data, sophisticated modeling can enhance diagnostic accuracy.

2. Cross-Frequency Coupling (CFC) Stimulation

Moving beyond traditional single-frequency stimulation, Kim et al. focused on Cross-Frequency Coupling (CFC). They utilized CFC-transcranial alternating current stimulation (CFC-tACS) to target the interactions between different frequency bands. This method directly addresses the functional mechanisms of the brain, leading to measurable improvements in cognitive performance.

3. Inter-Stimulus Synergy

The work of de Freitas Zanona and colleagues investigated the synergy between repetitive transcranial magnetic stimulation (rTMS) and sensory stimulation (SS). Their research highlights the importance of multi-modal approaches, showing how combined stimuli can affect brain regions that single-mode interventions cannot reach.

4. Oculomics and Biomarker Discovery

Expanding the search for biomarkers beyond the cranium, Guo et al. revealed connections between retinal microvascular changes and Neuromyelitis Optica Spectrum Disorder (NMOSD). This reinforces the potential of "oculomics"—using the eye as a non-invasive window into the brain's vascular and inflammatory health.

Key Insights: Synergy and Cognitive Mechanisms

The research presented in this collection offers several critical insights for the future of neuromodulation:

  • The Power of Combined Stimulation: A standout finding was that while rTMS and sensory stimulation individually altered somatosensory cortex (S1) excitability, only their combination successfully increased primary motor cortex (M1) excitability. This suggests that complex couplings between different sensory and motor pathways can be leveraged to unlock new therapeutic avenues, particularly in stroke rehabilitation.
  • Functional Coupling as a Lever: Cross-frequency coupling is shown to be a functional mechanism rather than a mere byproduct of neural activity. By specifically targeting these couplings, researchers were able to significantly reduce response times in working memory tasks, demonstrating a direct link between dynamic signal modulation and cognitive efficiency.
  • Interpretability Over "Black Boxes": While end-to-end deep learning models (like CNNs and attention models) are powerful tools for decoding, the authors emphasize that studying complex couplings as a whole is a vital alternative. For clinical practice, revealing the why behind a successful intervention is as important as the intervention itself.

Clinical Implications and Results

The practical applications of these theories are broad, ranging from diagnostic tools to performance enhancement:

  • Cognitive Enhancement: The use of CFC-tACS resulted in a significant reduction in response times during working memory tasks, pointing toward non-invasive ways to boost cognitive function.
  • Rehabilitation Pathways: The synergy between rTMS and sensory stimulation provides a potential roadmap for more effective stroke recovery protocols by targeting M1 excitability.
  • Diagnostic Accuracy: Advanced data augmentation techniques allow for more robust identification of brain disorders, even when faced with the "small data" problem common in medical research.

Future Directions: Optimization and Interpretability

Despite these advances, the field faces ongoing challenges. The editorial identifies an urgent need for the optimization of stimulation parameters. Currently, predicting how an individual patient will respond to neuromodulation remains difficult. Future research must focus on creating predictive models that account for individual variability in brain dynamics.

Furthermore, the transition from "black-box" AI to interpretable models remains a priority. To ensure the safety and efficacy of long-term clinical therapies, researchers must continue to develop methods that allow for a meaningful interpretation of the underlying brain dynamics. As we move forward, the integration of these diverse methodologies will be essential for turning complex neural signals into actionable clinical insights.

Summary updated

Abstract

The human brain is highly dynamic and complex, supporting a remarkable range of functions by dynamically integrating and coordinating different brain regions and networks across multiple spatial and temporal scales. Research on the human brain has become truly interdisciplinary involving medicine, neurobiology, engineering, and related fields. A thorough understanding of the mechanisms of neuromodulation actions is urgently needed for stimulation parameters optimization, response prediction, and consistent therapy. This Research Topic aims to combine top-down and bottom-up methods to produce robust results that allow for a meaningful interpretation in terms of the underlying brain dynamics with an emphasis on brain decoding and neuromodulation. Since the nonlinear, non-stationary, and complex couplings in brain activity, extremely rich information, including temporal, spatial, frequency, phase, and connectivity features, is embedded in every single measurement [1]. Many methods are dedicated to extracting specific features from the measurement. Even though more and more end-to-end deep models have been utilized for brain activity decoding, including convolutional neural networks, graphical neural networks, attention models, capsule networks, generative models, and so on, revealing the underlying mechanisms is essential for clinical practices [2], especially, neuromodulation. Hence, an important alternative is to study these features as a whole and study the complex couplings among a wide range of brain activity [3].The collection of articles in this Research Topic showcases the diversity of theoretical and empirical developments across a wide spectrum of brain dynamics research into complex couplings. Although this research topic only accepts four articles following the review process, it still covers a surprisingly wide range of approaches. Liu et al. studied the crossdomain data augmentation and showed that combining spatial-temporal features can improve the richness of generated data and contribute to the identification of brain disorders [4]; Kim et al. focused on the cross-frequency couplings (CFC) and the CFCtranscranial alternating current stimulation (CFC-tACS) was used to improve working memory performance and resulted in a significantly reduced response time [5]; de Freitas Zanona and co-authors studied inter-stimulus coupling and showed that the somatosensory cortex (S1) repetitive transcranial magnetic stimulation (rTMS) and sensory stimulation (SS) alone or in combination the S1 excitability was changed, but only their combination increased primary motor cortex (M1) excitability [6]; Guo et al. revealed the connections between retinal microvascular changes and NMOSD [7].In summary, this research topic highlights multiple methods for capturing brain dynamics and coupling analysis with high potential in a wide range of applications, such as brain disorder identification [4], improvement of working memory [5], treatment of stroke [6] and biomarker discovery [7], and so on. The brain dynamics and coupling analyses, especially, have far-reaching implications on neuromodulation.
BibTeX
@inproceedings{guo2023editorial,
  author = {Yuzhu Guo and Yang Li and Hua‐Liang Wei and Yifan Zhao},
  title = {{Editorial: New theories, models, and AI methods of brain dynamics, brain decoding and neuromodulation}},
  booktitle = {Frontiers in Neuroscience},
  year = {2023},
  doi = {10.3389/fnins.2023.1302505},
}

Loading related content...

Figure