Using Actual and Imagined Walking Related Desynchronization Features in a BCI

Area of Science:

• Neuroscience
• Rehabilitation Engineering
• Biomedical Engineering

Background:

• Brain-computer interface (BCI) research is expanding into motor rehabilitation, with a focus on upper body function.
• Gait rehabilitation is critical for stroke patients, but BCI applications are limited by challenges in recording electroencephalography (EEG) during gross movements.
• This study explores the feasibility of using walking-related brain signal features for BCI applications in gait rehabilitation.

Purpose of the Study:

• To investigate if a BCI can be developed using walking-related desynchronization features.
• To assess the impact of walking movement complexity on BCI classification performance.
• To evaluate the classification of EEG signals into 'walking' and 'no-walking' states.

Main Methods:

• Two BCI experiments were conducted with healthy subjects performing cued walking, complex walking (backward/adaptive), and imagined versions of these tasks.
• Electroencephalography (EEG) data was recorded during the tasks.
• Machine learning algorithms were used to classify EEG data into walking and non-walking states.

Main Results:

• Despite the automaticity of walking and recording difficulties, brain signals related to walking were classified rapidly and reliably.
• Classification performance was significantly higher for actual walking movements compared to imagined movements.
• No significant improvement in classification performance was observed for more complex walking tasks (backward/adaptive) versus simpler cued walking tasks.

Conclusions:

• The findings demonstrate the potential of using EEG-based BCIs for gait rehabilitation.
• BCI classification of walking is feasible and reliable, even under challenging recording conditions.
• Future research can focus on refining BCI paradigms for effective gait rehabilitation in clinical populations.