Multi-gesture drag-and-drop decoding in a 2D iBCI control task

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Intracortical brain-computer interfaces (iBCIs) enable control for individuals with tetraplegia.
• Current iBCIs are limited in decoding long-duration discrete movements like 'click-and-hold' or 'drag-and-drop'.

Purpose of the Study:

• To investigate neural activity during sustained hand gestures (1-4s) in the motor cortex.
• To develop and evaluate a novel 'latch decoder' for improved iBCI performance in discrete movements.

Main Methods:

• Recorded neural activity from the left precentral gyrus in two participants using the BrainGate2 system.
• Classified neural signals for multi-class gestures and binary (attempt/no-attempt) detection.
• Evaluated a novel latch decoder using isolated sustained gestures and a drag-and-drop task.

Main Results:

• Hand gesture discriminability decreased significantly for durations exceeding 1 second.
• The latch decoder substantially improved decoding accuracy compared to standard methods.
• Enhanced performance was observed for both isolated gestures and integrated 2D cursor control.

Conclusions:

• Sustained gesture attempts exhibit unique neurophysiologic patterns in the human motor cortex.
• The latch decoder offers a promising approach for intuitive iBCI control of consumer electronics.
• This advancement could expand iBCI capabilities for individuals with tetraplegia.