Cortical Control of a Prosthetic Arm for Self-feeding

The complexity of brain machine interface (BMI) design, the almost "sci-fi" impact of the potential end product, and the merging of neuroscience, neurosurgery, and bioengineering in these pursuits combine to make the BMI motor prosthetic field one of the most high profile and exciting areas of scientific research. Advances in this field are frequently published in some of the highest impact journals and are widely disseminated by the popular press. It is important in evaluating this field to look at each incremental step in motor prosthetic development beyond the "wow" factor to determine how much closer we are to developing a functioning prosthesis to help paralyzed patients. The quest to develop a thought-controlled motor prosthetic limb combines a number of difficult technical requirements. A brain signal with enough specificity and complexity to be trained and translated into multidimensional movements is necessary. The recorded neural activity must be maintained over time, without signal dropout due to glial scarring or loss of the recorded cells. The electronic technology will ideally need to be both miniaturized and also completely implantable, with remote transmission of information from the brain recording to the motor prosthesis. Finally, sensory feedback from the prosthetic limb should be able to scale the motor output through the prosthesis. The latest development in this field, published in Nature by Velliste et al. (453:1098–1101, 2008), represents a striking advance in one area of BMI motor neuroprosthetics. Researchers working at the University of Pittsburgh used intracortical microelectrode arrays implanted in the motor cortex of monkeys to record populations of single- and multiunit neuronal spiking activity. The authors translated the modulations in motor cortex neuronal activity that occurred when food was presented at different target locations to multidimensionally control a prosthetic arm. Amazingly, the monkeys rapidly learned how to control their own neuronal activity to feed themselves using the prosthetic arm. They were able to alter prosthetic trajectory when the food target was unexpectedly moved. They even learned novel behaviors with the prosthesis such as "finger" licking of the gripper portion of the prosthesis. This emergent behavior was not part of the experimental paradigm and represents embodied control of the prosthesis by the monkey. This work by Velliste et al. in the Schwartz lab provides compelling justification for the enthusiasm accompanying BMI motor prosthetic development. It represents the state-of-the-art in multidimensional prosthetic motor control. However, significant requirements remain in all areas of prosthetic design, including signal stability, remote signal transmission, electronics miniaturization, sensory feedback, and motor dexterity. For example, it remains unclear whether the powerful recordings of single- and multiunit activity, such as those used in this study, can be stably maintained for long-term prosthetic use. In contrast, surface electrocorticogram recordings have excellent stability over time, but may not provide enough signal resolution for multidimensional prosthetic control. BMI technology will continue to rapidly evolve. The development of human motor prosthetics, while not imminent, is inevitable. As neurosurgeons, we have a unique technical and scientific ability to work closely within multidisciplinary groups to advance and apply this technology to humans. GUY M. MCKHANN II, MD EPILEPSY AND NEUROTRANSPLANTATION RESEARCH