Clinical Instructor, Neurosurgery Stanford University Palo Alto, California, United States
Disclosure(s):
Matthew Willsey, MD, PhD: No financial relationships to disclose
Introduction: Intracortical brain-computer interfaces (iBCIs) may allow motor restoration for people with paralysis. The decoding algorithm, that converts neural activity to a control signal, is critically important for high-performing prosthetic devices. Herein, we adapt a high-performing feed-forward neural network decoding approach, used in able-bodied NHPs, for a human participant with paralysis to demonstrate simultaneous decoding of two finger groups.
Methods: A 69 yo man with C4 AIS C spinal cord injury had two 96-channel microelectrode arrays placed in the hand ‘knob’ area of left precentral gyrus 6 years before data collection. A monitor displayed a virtual hand with simultaneous movements of the thumb and index fingers, each in a one-dimensional arc. The neural network algorithm (NN) was trained having the participant attempt movements in sync with displayed fingers. During closed-loop control, the two fingers were moved simultaneously to acquire targets within their range of motion. Algorithm parameters were updated after closed-loop control based on the intended finger movements. The percent of successful trials and target acquisition times were used to evaluate closed-loop decoding. Pearson’s correlation (R) between the instructed movements and the NN-predicted movements was used to evaluate offline prediction.
Results: During closed-loop control (100 trials), the participant achieved a 100% success rate in acquiring targets for two independent, simultaneous finger targets with a mean acquisition time of 1220 ± 90 ms (mean ± S.E.M.). In 3 open-loop sessions, the Pearson’s correlation (R) between imagined and predicted finger velocities was high (0.72 ± 0.02).
Conclusion : This demonstrates that two finger movements, similar to those in able-bodied NHPs, can be easily decoded in a human participant using intracortical arrays in place for more than 6 years. The results have implications for fine motor control for iBCI as well as control of multiple effectors including reanimation of fingers, robotic finger movements, digital interfaces, and bimanual effectors.
