Multistep model for predicting upper-limb 3D isometric force application from pre-movement electrocorticographic...

Area of Science:

• Neuroscience
• Motor Control
• Brain-Computer Interfaces

Background:

• Neural signals in motor and non-motor regions may precede movement onset and encode direction.
• Electrocorticography (ECoG) offers high temporal and spatial resolution for studying neural dynamics.

Purpose of the Study:

• To investigate the presence and timing of neural correlates for movement planning onset and direction using ECoG.
• To develop and validate a computational model for decoding movement parameters from pre-movement neural activity.

Main Methods:

• Utilized a three-stage computational model: jPCA-RR-HMM, RDA, and LASSO regression.
• Analyzed human ECoG data during an upper-limb 3D isometric force task.
• Focused on pre-movement neural signals to predict movement onset and direction.

Main Results:

• Achieved a 60% true positive rate for force onset prediction within 250ms.
• Predicted one of six planned 3D movement directions with 36% accuracy (above chance).
• Identified direction-distinguishing information up to 400ms before movement onset, particularly in dorsal premotor regions.

Conclusions:

• The developed model effectively decodes movement onset and direction from pre-movement neural signals.
• This approach enables earlier and more accurate decoding of movement goals.
• Findings contribute to understanding the spatiotemporal dynamics of human motor planning and inform sensor placement for BCI applications.