Connecting Robot and Brain - Keynote to Bielefeld-Osaka Workshop 2010

12 October 2010
Begin time: 
End time: 

Brain-machine interface (BMI) is a technique that enables us to control
external devices and to communicate with others by brain signals alone.
We are presently developing a BMI system for functional restoration
using brain surface electrodes (electrocorticograms: ECoGs). ECoGs are
excellent regarding spatial resolution as well as signal-to-noise ratio,
and are notable for their ability to provide long term stable
recordings, which is an important factor for clinical BMI. It is
essential to understand what neurophysiological features provide better
neural decoding results. Using a support vector machine, we demonstrated
that ECoG signals from the brain groove called the central sulcus
provide higher decoding accuracy than those from brain outer surface.
Based on these findings, we developed a successive movement
classification estimating the decoding accuracy using mutual information
to provide smooth and robust control. We applied this method to an
ECoG-based real time control of a robot arm. A patient was able to hold
and release objects smoothly using the robot arm, which indicates that
our method is clinically feasible. We are also developing a
fully-implantable wireless system, which will be indispensable for
reducing the risk of infection in clinical applications. This system
includes a 3D high density electrode array, a multichannel analog
amplifier, a wireless battery/charger, and a wireless data transfer
circuit. Finally, neuroethical issues have to be considered carefully.
To summarize, an integrative approach bringing together neurophysiology,
computational neuroscience, medical engineering, robotics and
neuroethics is indispensable for the clinical application of BMI for
functional restoration.