A Bayesian dynamic stopping method for evoked response brain-computer interfacing.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interface (BCI) systems require enhanced speed, reliability, and user experience for broader applications beyond assistive technology.
• Current dynamic stopping methods in BCIs optimize metrics like symbols per minute (SPM) and information transfer rate (ITR), but may not suit all applications or users.
• Existing algorithms often rely on arbitrary thresholds and extensive training data, limiting their adaptability.

Purpose of the Study:

• To introduce a novel model-based dynamic stopping approach for BCIs.
• To enable precise control over error types and the balance between precision and speed in BCI systems.
• To offer a customizable solution for diverse BCI applications and user needs.

Main Methods:

• Developed a model-based dynamic stopping method leveraging analytical knowledge of the underlying classification model.
• Employed a risk minimization framework to precisely control error types and accuracy-speed trade-offs.
• Validated the proposed method using a publicly available BCI dataset.

Main Results:

• The proposed method demonstrated a broad range of accuracy-speed trade-offs.
• Achieved higher precision compared to established static and dynamic stopping methods.
• Showcased adaptability for customizing BCI performance.

Conclusions:

• The model-based approach offers superior performance and adaptability for BCI systems.
• This method provides precise control over BCI system performance, enhancing user experience and application suitability.
• Future BCI development can benefit from this risk-minimization strategy for optimized speed and accuracy.