Brain-machine interfaces, microstimulation, and movement primitives

Thumbnail

Event details

Date 03.06.2014
Hour 10:0011:00
Speaker Dr Simon Overduin, University of California, Berkeley USA
Location
Category Conferences - Seminars
Brain-machine interfaces (BMIs) offer the potential to recover movement plans from the brains of paralyzed individuals. Work over th¬e past few years has successfully used motor cortical activity to control computer cursors, robots, or muscle stimulators of up to a handful of independent dimensions. These demonstrations typically involve non-paralyzed, non-human animal models such as rhesus macaques. Translation of this work to human patients presents several major challenges, many of them intrinsic to the nature of intact animal models. For instance, decoders tested in non-paralyzed subjects typically rely on overt (rather than covert) movement control, intact (rather than interrupted) sensations, and (at most) short-term neural adaptation to the decoder. My work with neuroimaging and transcranial magnetic stimulation of human motor cortex highlights each of these limitations. However, the focus of my talk will be on limitations in the architecture of neuroprosthetic decoders themselves. As commonly tested in animal models, these decoders implicitly assume that motor control is updated moment-by-moment, concerned with extrinsic movement variables (such as Cartesian hand speed), and scalable to arbitrarily many dimensions. Yet in biological movement control, the nervous system arguably uses “motor primitives” to reduce the frequency at which commands are issued to groups of muscles acting in concert. Such primitives can be causally demonstrated by intracortical microstimulation of motor cortex. In macaques, microstimulation-evoked forelimb movements have smooth speed profiles characteristic of intermittent submovements seen in natural behaviors. The microstimulation-evoked forces tend to drive the arm towards invariant postures in the workspace around the animal. Across stimulation sites, the underlying muscle activations are reducible to linear sums of a few basic patterns, each corresponding to a muscle synergy evident in natural movements. I argue that the brain—and potentially BMIs—can simplify motor planning by exploiting these properties of primitives: intermittent planning, postural equivalence, and low dimensionality.

Bio:
Dr.  Overduin is a postdoctoral researcher working in the areas of motor control and neuroprosthetics, with Dr. Jose Carmena in the Department of Electrical Engineering and Computer Sciences at UC Berkeley. He received his PhD in Systems Neuroscience from MIT following studies of motor control and learning with Dr. Emilio Bizzi, and a BSc in Biology and Psychology from Wilfrid Laurier University in Canada. In his research, Dr. Overduin has aimed to understand and resolve some basic controversies in motor control and learning, including the necessity of primary motor cortex in early sensorimotor adaptation, the existence of motor consolidation in later sensorimotor learning, and the neurophysiological distinction between sensorimotor adaptation and sequence learning, as well as that between overt and covert movement. His current work involves translating the results of his neurophysiological studies into new experimental models and neuroprosthetic decoders for movement disorders including dystonia and spinal cord injury. Dr. Overduin has won fellowships and grants from numerous public and private research agencies in Canada and the US, including NSERC and CIHR in Canada, and the American Heart Association, Dystonia Medical Research Foundation, and NSF in the US.

Practical information

  • Informed public
  • Free

Organizer

  • Recruitment seminar CNP, IBI, BMI

Share