Finger movement and coactivation predicted from intracranial brain activity using extended block-term tensor...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Human intracranial finger movement recordings exhibit complex multilinear structures.
• Accurate decoding of these movements is crucial for developing effective brain-computer interfaces (BCIs).
• Existing regression methods may not fully capture the multilinear nature of neural data.

Purpose of the Study:

• Introduce extended Block-Term Tensor Regression (eBTTR), a novel regression method.
• Address the multilinear characteristics inherent in intracranial finger movement recordings.
• Improve the accuracy and efficiency of neural decoding for BCIs.

Main Methods:

• Developed eBTTR, a novel regression technique.
• Employed recursive Tucker decomposition for tensor analysis.
• Integrated automatic component extraction for feature selection.

Main Results:

• eBTTR demonstrated superior performance compared to state-of-the-art regression methods.
• The method accurately predicted intended finger trajectories.
• eBTTR also effectively predicted unintentional finger coactivations.

Conclusions:

• eBTTR offers a powerful and computationally efficient approach for decoding finger movements from intracranial recordings.
• Its performance rivals current leading methods, making it suitable for acute presurgical BCI development.
• The method's efficiency is advantageous given the time constraints in presurgical patient workups.