Motor cortical activity changes during neuroprosthetic-controlled object interaction

Abstract

Brain-computer interface (BCI) controlled prosthetic arms are being developed to restore function to people with upper-limb paralysis. This work provides an opportunity to analyze human cortical activity during complex tasks. Previously we observed that BCI control became more difficult during interactions with objects, although we did not quantify the neural origins of this phenomena. Here, we investigated how motor cortical activity changed in the presence of an object independently of the kinematics that were being generated using intracortical recordings from two people with tetraplegia. After identifying a population-wide increase in neural firing rates that corresponded with the hand being near an object, we developed an online scaling feature in the BCI system that operated without knowledge of the task. Online scaling increased the ability of two subjects to control the robotic arm when reaching to grasp and transport objects. This work suggests that neural representations of the environment, in this case the presence of an object, are strongly and consistently represented in motor cortex but can be accounted for to improve BCI performance.

Trial registration: ClinicalTrials.gov NCT01364480 NCT01894802.

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Figure 1

Neural response to object interaction task. Standardized firing rates from Subject 1 (top row) and Subject 2 (bottom row) averaged across trials from a single day of testing are shown for each recording channel. The channels are sorted according to the difference in standardized firing over the last 250 ms of the reach phase between the object and no object conditions with the units showing the highest difference at the top of each plot. On the testing day that is shown, Subject 1 completed 5 trials per condition and Subject 2 completed 50 trials per condition.

Figure 2

Population firing rate trajectories. The thin faded lines show the z-scored population firing rate for the 1.5 seconds before and after the hand reached the target (vertical dashed line) for every trial across all testing days (6 for Subject 1, and 2 for Subject 2). The bold line shows the average z-scored population firing rate for each condition. Both subjects showed increased population firing before object contact, while neither showed increasing firing rates while reaching towards the empty target region.

Figure 3

Movement trajectories near the target across all testing days. (a) The endpoint position of the arm is shown for the 1.5 seconds before (faded color) and after (saturated color) Subject 2 guided the hand to the target region. The left column shows reaches without scaling. The right column shows reaches with scaling. With scaling, the reach movement is much more consistent and there is less movement once the hand reaches the target. (b) The cumulative distribution of the path length in the 1.5 seconds after the hand reached the target and the subject was cued to grasp. Once the hand reaches the target no more arm movement is necessary, so longer paths (conditions that are further right in the plot) demonstrate less ability to stabilize the hand.

Figure 4

Object interaction and transport tasks. (a) The object interaction task required the subjects to move to a position target on the left side of the workspace before closing the hand. This task was repeated with and without and object in the workspace. (b) The object transport task required the subjects to grasp the cylindrical object on the left side of the table, carry it over the taped off region, and place it on the right side of the table as many times as possible within two minutes.