Characterization of the Sense of Agency over the Actions of Neural-machine Interface-operated Prostheses

True limb replacement requires blurring the cognitive perceptual lines between self and machine. Agency is the crucial bridge to perceiving the actions of an artificial limb as generated by self and intentional. This technique provides unique insights into the effectiveness of the cognitive perceptual communication between the human user and the machine.

This method could extend into any examination of the effects of feedback on agency, such as with myoelectric or body powered prostheses, cortical stimulation, stroke, or even hand transplants. Demonstrating this procedure is Dylan Beckler and Zachary Thumser, research engineers in our laboratory. Escort the participant into the testing room, and have them sit in front of a monitor.

Begin by generating a hand kinesthetic percept through the participants neural machine interface, or NMI, and capture the kinematics of the perceived motion by having the participant demonstrate what they feel using their intact hand. Use a virtual hand or prosthesis simulation to reproduce the kinematics of the movement percept. Next, set up hardware to capture the intentional hand movement control signals from the participant's NMI.

Map this control signal to the activity of the virtual prosthesis. Then, create a master control program that coordinates the acquisition of the NMI control signal, the movement of the virtual prosthesis, and of the generation of kinesthetic NMI feedback in real time. Display the virtual prosthesis on the monitor and adjust its size and location so that it is positioned congruently with the location of their missing limb.

Then, render objects, such as floating balls, in the virtual environment to service stop points for the close and open positions of the hand. Finally, configure the master control program so that when the virtual digits contact the virtual stop points, an auditory tone is played after an adjustable time delay. Prior to starting the experiment, build an input file for the master control program that specifies the settings for each trial, as well as the control and experimental conditions.

Program additional conditions designed to parse out the contributions to agency of motor intent, kinesthetic sensation, and temporal mismatch with the displayed kinematics of the virtual prosthesis. For example, consider using the following five conditions, opposite movement, too fast, too slow, onset delay, and no feedback. To initiate each trial, press the start button on the master control program, which moves the virtual hand to the start position, signaling the beginning of the trial.

In the first practice session, have the participant drive the hand to the movement end point and play the auditory tone 1, 000 milliseconds after the virtual digits reach the virtual stop point for 10 trials. In the second practice session, again have the participant drive the hand to the movement end point. Randomize the auditory tones so that the 300, 500, and 700 millisecond delay intervals are presented at least five times each.

Record the participants verbally reported estimation of the time delay interval. Finally, proceed with the experimental sets of 15 trials for each condition. Present the conditions in a randomized order, and administer a questionnaire at the end of each condition.

Here, results indicate the average score for the four agency questions, as well as four control questions for each participant, and by each feedback condition. An average rating greater than one indicates an agreement with a given statement, and zero indicates neutrality of agreement. Differences between the actual and perceived time intervals were then averaged across the three participants, and are presented relative to the baseline feedback condition.

Larger negative differences are an indication of a stronger implicit sense of agency, such as with the too fast condition. Further, these results show a comparison of explicit and implicit agency measures. The too fast condition demonstrated the strongest formation of agency, both explicitly and implicitly.

Minor modifications could expand our techniques to study additional sensor modalities, such as touch, and how they effect agency. Also, adding limb ownership measurements could help us understand the interrelationships between agency and embodiment. This technique enabled exploration of the role of perceptual movement feedback in a neural machine interface on the intrinsic mechanism that the brain uses to establish authorship over intended movements.