Single Neuron Control in Brain-Machine Interfaces
Garcia-Garcia, Martha 1, 2 ; Marquez-Chin, Cesar 2 ; Popovic, Milos 1, 2
1. Institute of Biomaterials and Biomedical Engineering, University of Toronto; 2. Toronto Rehabilitation Institute, University Health Network
Background: In the development of brain-machine interfaces (BMI), operant conditioning of neural activity is a strategy that has recently re-emerged from the need to design BMI systems that feel natural and intuitive to the user. In this strategy, a single cortical neuron is conditioned to produce an activity pattern that determines when the BMI is triggered.
Objectives: Compare performance differences between 2 types of cortical neurons undergoing a BMI task: fast-spiking, which produce fast trains or bursts, and regular-spiking, which produce slow trains of activity.
Design/Methods: A rat was implanted with an electrode array in the motor cortex and trained to trigger a reward dispenser by activating a single cortical neuron at progressively higher firing rates. Biofeedback of the firing rate was provided as the change in brightness of a light-emitting diode. The rat was trained in 10-20-minute long experiments at a time. Neurons were classified based on their firing behaviour and spike waveform shape.
Results: We found significant performance differences between fast-spiking neurons and regular-spiking neurons. Fast-spiking neurons were able to reliably produce the desired activity pattern within minutes of training, often immediately after trial onset. In contrast, regular-spiking neurons did not show any signs of being under the effect of operant conditioning of neural activity.
Conclusions: Improving the performance and usability of BMI systems has important implications for brain-controlled assistive devices, which are used on activities of daily living by the spinal cord injury population.