BIOMEDICAL ENGINEERING

Future generations of upper extremity neural prostheses using functional electrical stimulation (FES) will likely require control of several degrees of freedom. Promising sources of multi-degree-of-freedom command signals include unit activity from intracortical microelectrode arrays, field potentials recorded from the brain or scalp surface, and electromyographic (EMG) recordings from non-paralyzed muscles. Several studies have demonstrated these command signals can control a cursor or robot in two or three dimensions. It is assumed that these command sources could be used to control an arm/hand neuroprosthesis, but requirements of controlling arm movements under real-world conditions will be more demanding due to more degrees of freedom, arm and muscle dynamics, and limitations related to FES. This study helps define the required features of a brain-based command source for an upper extremity neuroprosthesis using a real-time computer simulation of an FES-enabled arm.
To ensure functional relevance, we identified specific functional tasks deemed crucial to an FES user, and then defined the kinematic requirements to accomplish these tasks. When using the computer simulation, a task is considered successful when a user stays within the kinematic limits for the entire task. This information defines the mechanical requirements of the FES system.
We also demonstrate a method to exploit natural synergies in the arm to determine the number of independent command signals needed to effectively and continuously command an FES-enabled arm. Rather than increase the number of command signals, we reduced the number of independent motions. This was accomplished using Principal Components Analysis (PCA), performed on the joint angles recorded from a wide range of motions performed by able-bodied subjects. This technique allows us to determine the minimum number of independent command signals needed to provide effective commands to the arm simulator.
Finally, we show how this PCA-based command method can be used to give commands to the real-time arm simulator to perform functional tasks in a virtual-reality environment. Both able-bodied and paralyzed individuals can use the simulator. Able bodied subjects can give commands using a standard mouse or joystick, which each provide independent two-dimensional commands. Paralyzed subjects can give commands using brain activity recorded from intra- or extra-cortical electrodes, and/or using muscle activity from EMG sensors.

Host: Professor Robert Kirsch