A Novel Online Action Observation-Based Brain-Computer Interface That Enhances Event-Related Desynchronization.

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) are emerging for stroke rehabilitation, primarily using motor imagery (MI).
• Limitations of MI-based BCIs include individual variability in event-related desynchronization (ERD) and low multi-class accuracy.
• Existing methods struggle to consistently improve rehabilitation outcomes.

Purpose of the Study:

• To introduce a novel online action observation (AO)-based BCI system for stroke rehabilitation.
• To investigate the efficacy of AO-based BCI in inducing and utilizing neural signals.
• To compare the performance of the AO-BCI with traditional AO methods.

Main Methods:

• Developed an AO-BCI using visual stimuli of four hand movements to elicit steady-state motion visual evoked potentials (SSMVEPs) and activate the sensorimotor region.
• Employed task-related component analysis for SSMVEP identification.
• Implemented a closed-loop system providing visual feedback based on SSMVEP detection.

Main Results:

• The amplitude of SSMVEP showed a negative relationship with stimulus frequency.
• Achieved a four-class classification accuracy of 72.81 ± 13.55% within 2.5 seconds.
• The AO-based closed-loop BCI significantly enhanced ERD compared to AO alone, potentially due to increased attentiveness.

Conclusions:

• The proposed AO-based BCI offers a new paradigm for stroke rehabilitation.
• This approach may overcome limitations associated with MI-based BCIs.
• AO-BCI demonstrates potential for improving neural plasticity and recovery in stroke patients.