Feasibility of Automatic Error Detect-and-Undo System in Human Intracortical Brain-Computer Interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Brain-computer interfaces (BCIs) translate neural signals into commands for assistive devices.
• Current BCIs require manual error correction, reducing efficiency and user satisfaction.
• Previous studies in monkeys indicated task-outcome signals in motor cortex activity.

Purpose of the Study:

• To investigate the presence of task-outcome signals in the human motor cortex.
• To determine if these signals can be accurately decoded to detect errors in real-time.
• To explore the potential for automated error detection and correction in BCIs.

Main Methods:

• Analysis of intracortical neural activity from two participants in the BrainGate2 trial.
• Participants controlled a computer cursor using neural signals for target selection and typing tasks.
• Posthoc analysis of action potentials and local field potentials.

Main Results:

• A putative error signal was identified in the human motor cortex's hand area, present in both spiking and local field potentials.
• Neural activity alone allowed for accurate classification of target selection outcomes (70-85% accuracy).
• High accuracy in detecting errors (0-3% misclassification of success trials) was achieved.

Conclusions:

• Human motor cortex contains neural signals indicative of task outcomes and errors.
• These signals can be decoded with high accuracy, enabling real-time error detection.
• Implementing an automated error detection and correction system could significantly enhance BCI performance and user experience.