Classifying motion states from neural activity of non-human primates for brain-computer interfaces

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interface (BCI) systems decode neural activity for effector control.
• Unintended effector activation occurs during intended non-movement due to persistent neural activity.
• Stable BCI control requires accurate identification of intended stationary states.

Purpose of the Study:

• To develop and evaluate a novel framework for classifying neural states in BCI.
• To distinguish between stationary and movement states directly from intracortical neural activity.
• To improve the stability and reliability of BCI control systems.

Main Methods:

• Proposed a neural-state classification framework (cpSVM) using principal component analysis, correlation-based feature selection, and a linear support vector machine.
• Utilized multi-unit recordings from premotor and primary motor cortices in non-human primates.
• Compared cpSVM performance against a conventional kinematics-based threshold-crossing method.

Main Results:

• Correlation-informed dimensionality reduction showed clear separation between stationary and movement states.
• The cpSVM achieved high classification accuracies (0.936 and 0.930).
• cpSVM outperformed the threshold-crossing method in accuracy, sensitivity, specificity, F-score, and output continuity.

Conclusions:

• Stationary and movement states can be reliably distinguished using a low-dimensional, correlation-informed classification approach.
• The cpSVM framework effectively suppresses unintended effector activation.
• This method enhances continuity and stability in BCI control systems.