Error detection and correction in intracortical brain-machine interfaces controlling two finger groups.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-machine interfaces (BMIs) offer potential for restoring motor function but are limited by decoding errors.
• Existing error correction methods focus on post-trial outcomes, not real-time movement execution.
• Continuous detection and correction of execution errors are crucial for improving BMI effectiveness.

Purpose of the Study:

• To develop and evaluate a method for real-time detection and correction of brain-machine interface (BMI) execution errors.
• To investigate the neural correlates of movement errors during BMI control.
• To assess the impact of error correction on overall BMI task performance.

Main Methods:

• Two male rhesus macaques were implanted with motor cortex Utah arrays for neural recording.
• A Kalman filter decoded neural activity into intended finger kinematics for BMI control.
• A novel method analyzed neural activity to detect movements deviating from target goals, enabling error correction via a stopping strategy.

Main Results:

• Neural activity variance was significantly explained by including target distance, alongside kinematics.
• The study demonstrates, for the first time, the online detection of BMI execution errors using motor cortex neural activity.
• A mean true positive detection rate of 28.1% was achieved with a <5% false positive rate, leading to improved task performance by reducing finger orbiting time.

Conclusions:

• Motor cortex neural activity can be leveraged to detect and correct BMI execution errors in real-time.
• The developed error detection and correction strategy significantly enhances BMI task performance.
• Future research should focus on refining classification and correction strategies for further performance gains.