Editorial: Exploration of the non-invasive brain-computer interface and neurorehabilitation

Shugeng Chen, Lin Yao, Lei Cao, Marco Caimmi, Jie JiaFrontiers in Neuroscience 2024 · DOI · OpenAlex

Abstract

Non-invasive Brain-Computer Interfaces (BCIs) represent a transformative frontier in neurorehabilitation, particularly for stroke recovery. This editorial synthesizes eight research contributions that address critical bottlenecks in BCI technology, including signal decoding accuracy, system complexity, and clinical translation. By integrating deep learning architectures, optimization algorithms, and longitudinal clinical studies, these works provide a roadmap for moving BCIs from laboratory settings to bedside applications.

Introduction

The integration of non-invasive BCIs, primarily based on electroencephalogram (EEG) signals, offers a promising pathway for restoring motor function in patients with neurological impairments. However, the transition from experimental prototypes to standard clinical tools is hindered by several "bottlenecks." These include the inherent difficulty of decoding noisy brain signals during complex motor tasks, the impracticality of high-density electrode arrays in clinical environments, and the persistent gap between algorithmic performance and functional patient recovery. This Research Topic aims to bridge these gaps through innovation in algorithms, paradigm optimization, and the study of neural plasticity.

Methodological Advancements

Deep Learning and Transformers

Recent studies have moved beyond traditional machine learning, employing Local and Global Convolutional Neural Networks (CNNs) paired with self-attention transformers and sliding window attention mechanisms. These architectures are adept at extracting rich spatial-spectrum-temporal features from EEG data, significantly enhancing the decoding of motor intent.

System Optimization and Portability

To address the "complexity bottleneck," researchers utilized multi-objective evolutionary algorithms for channel selection. This approach identifies the minimal electrode set required for high-accuracy performance, facilitating the development of portable, user-friendly systems. Furthermore, the development of functional-oriented portable BCI equipment has shown both improved clinical outcomes and enhanced classification accuracy in real-world settings.

Calibration and Transfer Learning

The "calibration problem"—the need for long setup times for new users—is addressed through dual selections-based knowledge transfer. This allows systems to leverage data from previous subjects to initialize models for new patients, reducing the barrier to entry for clinical use.

Clinical Insights and Biomarkers

Stroke Recovery and Brain Plasticity

Clinical cohorts involving both subacute and chronic stroke patients provided evidence for BCI-driven recovery. Notably, action observation-driven BCI robotic training showed significant promise in 16 chronic stroke patients. Research identified specific biomarkers, such as ipsilesional oscillation changes in delta and beta bands, which correlate with recovery progress.

Connectivity Dynamics

Connectivity analysis revealed that subacute stroke patients exhibit higher transfer efficiency in brain networks compared to chronic patients. This suggests that the timing of BCI intervention relative to the injury is a critical factor in leveraging neuroplasticity for recovery.

Results and Discussion

Functional Gains

Significant improvements were recorded in standard clinical metrics, including the Action Research Arm Test (ARAT) and the Fugl-Meyer Assessment (FMA), following BCI-robotic training. These results underscore the potential for BCIs to drive meaningful functional recovery.

Algorithmic Performance

The introduction of transformers and attention mechanisms led to classification accuracies that are comparable or superior to previous state-of-the-art models. This suggests that deep learning is effectively overcoming the accuracy bottleneck in EEG decoding.

Conclusion and Future Directions

While progress is substantial, the field must transition from offline validation to real-time (online) control experiments to demonstrate active motor control capabilities. Future efforts should prioritize hardware robustness for home use and the execution of multi-center randomized controlled trials to establish high-level clinical evidence. Closing the gap between laboratory paradigms and clinical requirements remains the primary objective for the next generation of neurorehabilitation technology.

Summary updated

Abstract

Keywords: brain-computer interface (BCI), electroencephalogram (EEG), stroke, rehabilitation, algorithm
BibTeX
@inproceedings{chen2024editorial,
  author = {Shugeng Chen and Lin Yao and Lei Cao and Marco Caimmi and Jie Jia},
  title = {{Editorial: Exploration of the non-invasive brain-computer interface and neurorehabilitation}},
  booktitle = {Frontiers in Neuroscience},
  year = {2024},
  doi = {10.3389/fnins.2024.1377665},
}
Figure