Learning to Control a Brain–Machine Interface for Reaching and Grasping by Primates

Received August Learning to Control a Brain-Machine Interface for Reaching and Grasping by Primates Jose M. Carmena 0 Mikhail A. Lebedev 0 Roy E. Crist 0 Joseph E. O'Doherty 0 David M. Santucci 0 Dragan F. Dimitrov 0 Parag G. Patil 0 Craig S. Henriquez 0 Miguel A. L. Nicolelis 0 Idan Segev, Hebrew University 0 1 Department of Neurobiology, Duke University , Durham , North Carolina, United States of America, 2 Department of Biomedical Engineering, Duke University , Durham , North Carolina, United States of America, 3 Division of Neurosurgery, Duke University , Durham , North Carolina, United States of America, 4 Center for Neuroengineering, Duke University , Durham , North Carolina, United States of America, 5 Department of Psychological and Brain Sciences, Duke University , Durham, North Carolina , United States of America Reaching and grasping in primates depend on the coordination of neural activity in large frontoparietal ensembles. Here we demonstrate that primates can learn to reach and grasp virtual objects by controlling a robot arm through a closed-loop brain-machine interface (BMIc) that uses multiple mathematical models to extract several motor parameters (i.e., hand position, velocity, gripping force, and the EMGs of multiple arm muscles) from the electrical activity of frontoparietal neuronal ensembles. As single neurons typically contribute to the encoding of several motor parameters, we observed that high BMIc accuracy required recording from large neuronal ensembles. Continuous BMIc operation by monkeys led to significant improvements in both model predictions and behavioral performance. Using visual feedback, monkeys succeeded in producing robot reach-and-grasp movements even when their arms did not move. Learning to operate the BMIc was paralleled by functional reorganization in multiple cortical areas, suggesting that the dynamic properties of the BMIc were incorporated into motor and sensory cortical representations. - Traumatic lesions of the central nervous system as well as neurodegenerative disorders continue to inflict devastating, and so far irreparable, motor deficits in large numbers of patients. Every year, spinal cord injuries alone are responsible for the occurrence of about 11,000 new cases of permanent paralysis in the United States (Nobunaga et al. 1999). These cases add up to an already sizeable population of patients, estimated at 200,000 in the United States (Nobunaga et al. 1999), who have to cope with partial (as in the case of paraplegics) or almost total (i.e., quadriplegia) body paralysis. Until very recently, the main thrust of basic research on restoration of motor functions after spinal cord injuries focused on reconstructing the connectivity and functionality of damaged nerve fibers (Ramon-Cueto et al. 1998; Uchida et al. 2000; Bomze et al. 2001; Bunge 2001; Schwab 2002). While this repair strategy has produced encouraging results, such as limited restoration of limb mobility in animals, the goal of restoring complex motor behaviors, such as reaching and grasping, remains a major challenge. Two decades ago, an alternative method for restoring motor behaviors in severely paralyzed patients was proposed (Schmidt 1980). This approach contends that direct interfaces between spared cortical or subcortical motor centers and artificial actuators could be employed to bypass spinal cord injuries so that paralyzed patients could enact their voluntary motor intentions. Initial experimental support for a cortically driven bypass came from the studies conducted by Fetz and collaborators (Fetz 1969; Fetz and Finocchio 1971, 1975; Fetz and Baker 1973), who demonstrated that macaque monkeys could learn to selectively adjust the firing rate of individual cortical neurons to attain a particular level of cell activity if provided with sensory feedback that signaled the level of neuronal firing. Recent studies in rodents (Chapin et al. 1999; Talwar et al. 2002), primates (Wessberg et al. 2000; Serruya et al. 2002; Taylor et al. 2002), and human subjects (Birbaumer 1999) have rekindled interest in using brainmachine interfaces (BMIs) as a potential alternative for spinal cord rehabilitation. These experiments have demonstrated that animals can learn to utilize their brain activity to control the displacements of computer cursors (Serruya et al. 2002; Taylor et al. 2002) or one-dimensional (1D) to three-dimensional (3D) movements of simple and elaborate robot arms (Chapin et al. 1999; Wessberg et al. 2000). Despite these initial results, several fundamental issues regarding the operation of BMIs, ranging from basic electrophysiological issues to multiple engineering bottlenecks, remain a matter of considerable debate (Nicolelis 2001, 2003; Donoghue 2002). For example, although most agree that a BMI designed to reproduce arm/hand movements will require long-term and stable recordings from cortical (or subcortical) neurons through chronically implanted electrode arrays (Nicoleli (...truncated)