Cortical Decoding of Individual Finger Group Motions Using ReFIT Kalman Filter

Area of Science:

• Neuroscience
• Biomedical Engineering
• Robotics

Background:

• Brain-machine interfaces (BMIs) have advanced upper limb prosthetics.
• Precise finger control decoding remains a significant challenge.
• Existing BMI studies have limited focus on individual finger movements.

Purpose of the Study:

• To investigate online control of individual finger groups using a novel training manipulandum.
• To evaluate the performance of the ReFIT Kalman filter for precise finger decoding.
• To compare ReFIT Kalman filter with standard Kalman filter for finger control.

Main Methods:

• Developed a novel training manipulandum for isolating index and middle-ring-pinkie (MRP) finger movements in non-human primates (NHPs).
• Utilized the Recalibrated Feedback Intention-Trained (ReFIT) Kalman filter for decoding finger positions.
• Recorded motor cortex signals using Utah arrays in two rhesus macaques.

Main Results:

• NHPs successfully acquired fingertip targets using isolated index and MRP fingers.
• The ReFIT Kalman filter demonstrated a 31.0% and 35.2% improvement in bit rate over the standard Kalman filter.
• Decoding accuracy for differentiating index and MRP finger movements was 81.7%, but linear decoding of both simultaneously was insufficient.

Conclusions:

• This study presents the first systematic separation of digits for continuous online decoding in NHPs.
• The ReFIT Kalman filter enhances precise finger decoding performance in BMIs.
• Novel nonlinear approaches may be required for independent and precise individual finger control.