Facilitating Knowledge Transfer: An Approach for Matching Neural Patterns between Motor Tasks.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-machine interfaces (BMIs) offer potential for motor function restoration.
• Adapting BMIs to new tasks requires efficient knowledge transfer of neural-action mappings.
• Current BMI systems face challenges in real-world adaptability and rapid learning.

Purpose of the Study:

• To develop and validate a method for matching neural patterns between motor tasks to facilitate knowledge transfer in BMIs.
• To improve the efficiency and accuracy of adapting BMI decoders to new tasks.

Main Methods:

• Projecting neural data into a 6-dimensional (6D) joint-todimensional principal component analysis (jPCA) feature space.
• Identifying preserved neural patterns in the last four dimensions of the jPCA space.
• Utilizing a decoder trained on a previous task to estimate a matched class, refined by clustering in the first two jPCA dimensions.

Main Results:

• The proposed method achieved 87.04% accuracy in estimating the matched class between sequential motor tasks in a rat model.
• Using the entire jPCA space for decoder transfer resulted in significantly lower accuracy (39.8%).
• Neural patterns associated with the same action were observed to be preserved in specific dimensions of the jPCA space.

Conclusions:

• The developed approach effectively matches neural patterns across motor tasks, enabling efficient knowledge transfer for BMIs.
• This method enhances BMI adaptability to new tasks by leveraging previously learned neural-action mappings.
• The findings demonstrate the potential of targeted feature space analysis for improving BMI performance and user adaptation.