Abstract—The purpose of this study was to develop an animal model to evaluate the efficacy of a brain machine interface (BMI) to control a neuroprosthesis intended to restore hand function via functional neuromuscular stimulation (FNS). We have implemented the system in a single primate, whose limb could be temporarily paralyzed by a reversible peripheral nerve block. Recordings from the primary motor cortex were obtained from a 100-electrode array in the intact monkey, and used to predict the activity of a variety of wrist and hand muscles. These predictions were calculated in real-time, and used as inputs to a 4 channel neuromuscular stimulator for electrically activating the paralyzed muscles. Here we demonstrate that the BMI can be used to restore voluntary control of wrist flexion following muscle paralysis. I. INTRODUCTION UNCTIONAL neuromuscular stimulation has proven to be an effective means for restoring movement control following spinal cord injury [1]. Most of systems have focused on stance, bowel or bladder control, or basic grasping movements using low dimensional control signals. However, the restoration of more complex tasks, such as dexterous object manipulation, 3-dimensional reaching, or inter-limb coordination would require access to more control sources than are currently unavailable from severely disabled patients. Ironically, it is most difficult to obtain such control sources from the individuals who are most likely to benefit from them. For example, patients with spinal cord injuries at the C4 level or above are generally paralyzed from the shoulders down and would need restoration of full arm and hand movement to achieve functional reach and grasp [2]. Unfortunately, these patients have fewer options for command signals as only the head and neck area remain under volitional control. Manuscript received February 4, 2007. This work was supported by NIH NS053603. E.A. Pohlmeyer is with the Dept. of Biomedical Eng., Northwestern University, Chicago, IL 60611 USA. (e-pohlmeyer@northwestern.edu). E.J. Perreault is with the Depts. of Biomedical Eng. and Physical Med. and Rehab., Northwestern University, Chicago, IL 60611 USA. (e-perreault@northwestern.edu). M.W. Slutzky is with the Depts. of Physiology, Neurology, and Physical Med. and Rehab., Northwestern University, Chicago, IL 60611 USA. (mslutzky@northwestern.edu). K.L. Kilgore is with the Dept. of Orthopaedics, MetroHealth Medical Center, Cleveland, OH 44109 USA. (klk4@case.edu). R.F. Kirsch is with the Dept. of Biomedical Eng., Case Western Reserve University, Cleveland, OH 44106 USA. (rfk3@cwru.edu). D.M. Taylor is with the Dept. of Biomedical Eng., Case Western Reserve University, Cleveland, OH 44106 USA. (dxt42@cwru.edu). L.E. Miller is with the Depts. of Physiology and Biomedical Eng., Northwestern University, Chicago, IL 60611 USA. (lm@northwestern.edu; 1.312.503.8677). BMIs provide one possibility for providing higher dimensional control sources for FNS applications. The goal of this project is to test the efficacy of a BMI for controlling an FNS system. To accomplish this goal, we have developed a non-human primate model that includes an intracortical micro-electrode array, a reversible peripheral nerve block, and an FNS system to activate several muscles of the wrist and hand. This paper presents our initial evaluation of this system’s performance in restoration of the monkey’s ability to control voluntary wrist flexion force despite the temporary paralysis. II. METHODS A. Behavioral Tasks A single, male Rhesus monkey was trained to perform a variety of tasks that required either grasp force (including palmar, lateral, and precision grasps) or isometric wrist flexion/extension forces. Grasp force was measured using a pair of force sensitive resistors mounted on each of the devices. Wrist force was measured by a pair of strain gauges mounted on an aluminum cantilever strapped to the monkey’s arm. The monkey faced a video monitor on which a moving cursor provided visual feedback of the applied forces. Targets were also displayed on the monitor, appearing pseudo-randomly at one of several different levels. Under normal conditions, the monkey was required to move the cursor into the target within 1.5 seconds in order to receive a liquid reward. During nerve block, this period was increased to 3.0 seconds. A trial consisted of matching a single target, and the monkey acquired 1000 or more such targets in a typical session. Data files were collected while the monkey reached toward and grasped a series of the devices in random order. In addition to the standard force tracking paradigm, we also measured the maximum amount of force that the monkey would generate voluntarily (MVC) for each device. This estimate was important in order to gauge both the depth of the peripheral nerve block, and the efficacy of the FNS. Unlike similar paradigms with cooperative human subjects, this required the use of a non-monotonic sequence of increasing target force levels, by which means the monkey was coaxed to exert increasingly high force without giving up. This method typically proved capable of generating consistent estimates of MVC within 3-4 minutes. B. Surgery After the animal had become familiar with the tasks, Real-Time Control of the Hand by Intracortically Controlled Functional Neuromuscular Stimulation Eric A. Pohlmeyer, Eric J. Perreault, Marc W. Slutzky, Kevin L. Kilgore, Robert F. Kirsch, Dawn M. Taylor and Lee E. Miller F