A modular neural interface for massively parallel recording and control : subsystem design considerations for research and clinical applications

摘要

The closed-loop Brain-Machine Interface (BMI) has long been a dream for clinicians and neuroscience researchers alike - that is, the ability to extract meaningful information from the brain, perform computation on this information, and selectively perturb neural dynamics in the brain for therapeutic benefit to the patient. Such systems have immediate application to treatment of paralysis, epilepsy and the amputated, and the potential for treatment of higher order cognitive dysfunction. Despite the promise of the BMI concept, the technology for bidirectional communication with the brain at sufficiently large scale to be truly therapeutically useful is lacking. Current state-of-the-art neuromodulation systems deliver open loop, 16-channel patterned electrical stimulation incapable of precisely targeting small numbers of neurons. Large-scale neural recording systems are limited to 16-128 electrodes, at the cost of several thousand dollars per channel. The ability to record from the awake behaving animal - let alone precisely modulate neural network dynamics in closed-loop fashion- presents a substantial challenge today.