Surfing beta burst waveforms to improve motor imagery-based BCI

Area of Science:

• Neuroscience
• Brain-Computer Interfaces (BCI)
• Signal Processing

Background:

• Motor-related brain activity is often characterized by event-related desynchronization/synchronization (ERD/ERS) in mu and beta bands.
• Brain-computer interfaces (BCIs) utilize motor imagery (MI) paradigms, expecting ERD/ERS-like modulations.
• Recent studies suggest beta band activity occurs as transient bursts, not sustained oscillations, challenging traditional BCI assumptions.

Purpose of the Study:

• To develop and validate a signal processing algorithm for BCI applications based on beta bursts.
• To compare the efficacy of beta burst analysis against traditional beta band power for motor imagery classification.
• To explore motor imagery dynamics using a time-resolved decoding approach with novel classification features.

Main Methods:

• A novel signal processing pipeline was developed, filtering brain recordings using kernels derived from beta bursts.
• Spatial filtering was applied to the data before classification.
• A time-resolved decoding approach was employed to analyze MI dynamics and feature specificity.

Main Results:

• The proposed beta burst-based filtering and classification method demonstrated superior performance compared to traditional beta band power analysis.
• The algorithm is efficient, suitable for online applications, and compatible with state-of-the-art techniques.
• Classification performance and information transfer rates were improved by utilizing beta bursts.

Conclusions:

• Beta burst analysis offers a more effective approach for motor imagery classification in BCIs than conventional beta band power.
• The developed data-driven filtering pipeline provides a robust and efficient method for analyzing multi-sensor brain recordings.
• This approach has the potential to advance the capabilities of non-invasive BCIs for motor-related applications.