EFRI-M3C: Development of New Algorithmic Models and Tools to Enhance Neural Adaptation in Brain Computer Interface Systems.

Washington University, Saint Louis MO

Investigators

Abstract

Human interaction with machines has always relied on some form of muscle movement to translate the brain's desired action to the machine (e.g. turning a knob). The objective of our project is to eliminate the need for muscle transformations in the man-machine interface. Using a new brain-computer interface (BCI) technology (electrocorticography or ECoG) pioneered by the research team, we will develop novel decoding algorithms to control the force/torque inputs to an external device directly. Likewise, by designing machine learning algorithms to identify and incorporate neural plasticity in the decoding schemes will allow the BCI to evolve over time. Finally, by combining brain signals from multiple areas to identify various brain states, we can identify and change the effectors to be controlled. Intellectual Merit: All previous BCI studies decoded only kinematic signals to control computer cursors as well as robotic limbs. While kinematic control is a natural extension for disabled individuals trying to regain function lost by paralysis or amputation, a direct interface between the brain and the machine allows for much more elegant interaction. Thus, rather than controlling a robotic arm through the use of imagined self limb movements (a proxy of intention), one rather controls the device as if it was a part of their own body. For instance, mapping brain activity to kinetic parameters such as a robot?s torque motor allows the individual to directly control the forceful interactions within the system. Broader Impact: One advantage of ECoG is that while it is an invasive recording technology, the electrodes can be placed epidurally which significantly reduces the risk profile for implantation. In the long term, this should allow ECoG-based BCIs to be accepted as a viable implant in able-bodied humans. Developing a safe and effective BCI modality for the general public will fundamentally change how humans interact with machines. No longer will humans require muscle activity as an intermediary to interact with machines. A whole new field of man-machine interfacing will be initiated where the brain builds complex internal models of the machine's dynamics (instead of musculoskeletal dynamics) for accurate and direct control of the machine's effectors.