C5: A brain-machine interface for reaching and grasping based on intracranially implanted electrodes in humans
Tonio BallI, Carsten MehringJ and Martin NawrotQ
I = Epilepsy Centre, University Medical Center
J = Animal Physiol. & Neurobiol, Inst. of Biol.I
Q = Bernstein Center for Computational Neuroscience, Berlin
Scientific background
In rats and monkeys, activity of multiple single neurons related to paw resp. arm movement can be employed to control an external actuator. Based on this pioneering work, there is increasing interest in creating an implantable brain-machine interface (BMI), enabling real-time control of neuroprosthetic devices. Such movement inference has not yet been demonstrated in humans, and little is known about the possibility to decode information for the control of grasping from different sensorimotor areas activated by hand movements. Long-term stability of the underlying signals is essential for brain-machine interface applications. Thus, a promising new approach for robust neuro-interfacing is based on neuronal population activity, instead of multiple single neuron activity. We recently demonstrated that local field potentials can be as efficient as single neuron activity for decoding arm movements. In an ongoing project we have successfully shown the feasibility of inferring the direction of center-out reaching movements in humans from single-trial activity measured with electrodes on the cortical surface.
Objectives
The project aims at a better understanding of how the human brain controls voluntary movement and at developing a brain-machine interface for reaching and grasping in humans. We record human field potentials in epilepsy patients simultaneously from the cortical surface of different frontal and parietal sensorimotor areas with electrodes subdurally implanted for clinical evaluation. These recordings allow for a uniquely “close look” at neuronal population activity of the human brain underlying motor control. Neural dynamics related to different aspects of grasping (gripping, holding and releasing objects, grip force, precision grip) are studied and methods for extraction of signals for grasp control developed. Based on our recent findings, cortical potentials are decoded in real-time by advanced data analysis techniques to enable subjects to directly brain-control the reaching movement of an external actuator (computer cursor, robotic arm) which is visible to the subject. In particular, we are interested in the adaptivity of neuronal population activity during task learning within experimental sessions. Finally, real-time decoding of reaching and grasping is integrated in a brain-machine interface for simultaneous brain-control of grip and arm movement.