Sequential Probability Ratio Testing with Power Projective Base Method Improves Decision-Making for BCI.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Fast and reliable decision-making is critical for real-time brain-computer interface (BCI) applications, such as controlling wheelchairs or neuroprosthetics.
• Existing BCI methods face challenges in achieving optimal speed and accuracy for single-trial classification of motor imagery events.

Purpose of the Study:

• To develop and evaluate a novel decision-making model, sequential probability ratio testing (SPRT), for enhanced single-trial classification of motor imagery in BCI.
• To assess the performance of the SPRT method against benchmark techniques using EEG data from multiple subjects and datasets.

Main Methods:

• EEG signals were analyzed using a power projective base method.
• The sequential probability ratio testing (SPRT) model was applied for single-trial classification of motor imagery movement events.
• Performance was evaluated across thirteen subjects using three distinct datasets, comparing SPRT with benchmark and non-sequential methods.

Main Results:

• The proposed power projective method, integrated with SPRT, outperformed two benchmark methods across all subjects.
• Sequential classification using SPRT achieved significantly higher accuracies compared to non-sequential approaches.
• The SPRT method demonstrated an average maximum accuracy of 84.1%, surpassing the sequential Bayesian method (82.3%).

Conclusions:

• The SPRT method offers an accumulative process that enhances discriminative power over time, leading to improved BCI decision-making.
• SPRT provides a clear framework for balancing the trade-off between decision time and classification error.
• The findings suggest SPRT is a valuable tool for accelerating decision-making in BCI systems while managing error rates.