Single Finger Trajectory Prediction From Intracranial Brain Activity Using Block-Term Tensor Regression With Fast and...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Conventional vector- or matrix-based decoding for brain-computer interfaces (BCIs) often fail to capture the multilinear structure of brain signals.
• Existing tensor-based decoding methods show promise but are computationally intensive, limiting their practical application.

Purpose of the Study:

• To introduce a computationally efficient tensor-based decoding algorithm for BCIs.
• To improve the accuracy of predicting neural signals by leveraging the multilinear structure of brain data.

Main Methods:

• Developed two novel tensor factorizations integrated into the block-term tensor regression (BTTR) algorithm.
• Introduced a marginalization procedure for robust predictions and reduced overfitting (generalized regression).
• Applied BTTR to predict single finger movement trajectories from intracranial recordings in human subjects.

Main Results:

• BTTR demonstrated significant performance gains over conventional vector- and matrix-based techniques.
• The algorithm automatically accounts for underlying data structure in a computationally efficient manner.
• Achieved accurate prediction of single finger movement trajectories from intracranial recordings.

Conclusions:

• The proposed BTTR algorithm offers a computationally efficient and accurate approach for tensor-based decoding in BCIs.
• BTTR effectively handles the multilinear structure of brain signals, outperforming state-of-the-art methods in a real-world application.
• This advancement facilitates more robust and reliable brain-computer interface performance.