Continuous neuronal ensemble control of simulated arm reaching by a human with tetraplegia

Abstract

Functional electrical stimulation (FES), the coordinated electrical activation of multiple muscles, has been used to restore arm and hand function in people with paralysis. User interfaces for such systems typically derive commands from mechanically unrelated parts of the body with retained volitional control, and are unnatural and unable to simultaneously command the various joints of the arm. Neural interface systems, based on spiking intracortical signals recorded from the arm area of motor cortex, have shown the ability to control computer cursors, robotic arms and individual muscles in intact non-human primates. Such neural interface systems may thus offer a more natural source of commands for restoring dexterous movements via FES. However, the ability to use decoded neural signals to control the complex mechanical dynamics of a reanimated human limb, rather than the kinematics of a computer mouse, has not been demonstrated. This study demonstrates the ability of an individual with long-standing tetraplegia to use cortical neuron recordings to command the real-time movements of a simulated dynamic arm. This virtual arm replicates the dynamics associated with arm mass and muscle contractile properties, as well as those of an FES feedback controller that converts user commands into the required muscle activation patterns. An individual with long-standing tetraplegia was thus able to control a virtual, two-joint, dynamic arm in real time using commands derived from an existing human intracortical interface technology. These results show the feasibility of combining such an intracortical interface with existing FES systems to provide a high-performance, natural system for restoring arm and hand function in individuals with extensive paralysis.

Figure 1

Overview of experimental setup. The participant is seated in front of a computer monitor on which the virtual arm is displayed. Neural spike data are sorted from MI multielectrode recordings from the BrainGate array and passed into the decoder as spike occurrence times. Movement velocity is then decoded from ensemble patterns using a Kalman filter that was characterised during participant training. The FES controller then calculates the muscle activations required to produce the desired movement based on the decoded velocity command and feedback of the current position of the virtual arm (which would be provided by sensors in a real implementation). A real-time, dynamic arm model simulates the effect of the muscle activations on the dynamic movements of the virtual arm, and the resulting movement is displayed to the user via the computer monitor, thus completing the control loop for the user. The task of the participant is to use the virtual dynamic arm to acquire a series of targets presented on the screen.

Figure 2

Processing of raw neural data into velocity commands, muscle activation patterns, and virtual arm motions. For each trial, a series of targets were presented to the subject over a period of several minutes. 120s of data from one trial is shown to demonstrate data processing. (a) Spike events from 40 sorted units simultaneously recorded by the BrainGate array. (b) Decoded endpoint velocities estimated from the spike sorting and Kalman filter. (c) Activations of mono-articular (deltoid posterior - DP, deltoid anterior - DA, brachialis - BR, triceps brevis - TB) and bi-articular (triceps longus - TL and biceps longus - BL) muscles required to achieve the desired movement as calculated by the FES controller. (d) The x-y locations of the targets presented and the corresponding x-y locations of the finger as the arm moves towards those targets.

Figure 3

Summary of target reaching task performance. Target reaching tasks were completed in three sessions across three days for both the DAS and NO DAS conditions (“DAS” included arm and muscle dynamics, as well as the simulated FES controller; “NO DAS” did not include limb dynamics and was purely kinematic control similar to a 2D cursor control task). Performance was assessed using measures of average speed, path efficiency, and data throughput. The box plots have lines at the lower quartile, median, and upper quartile values, and whiskers extending to 1.5 times the interquartile range from the ends of the box. Values outside the whiskers are shown as crosses, and the overlapping notches of the box plots indicate no significant difference between data sets. No significant differences in task performance within a session were seen between the DAS and NO DAS conditions for the measures of path efficiency and throughput (rows one and three). Differences were seen between those two groups in the average speed of movement (row two) for sessions one (p=0.049) and two (p=0.0033), where the average speed of movement was higher for the kinematic only task.