Michigan State University - Research Groups

Dr. Oweiss spoke on September 8^th, 2010 at University College London, United Kingdom. His plenary talk, entitled "An engineer’s view of the brain: can we electronically read and write to the mind?", was part of INSPIRE 2010 Conference on information representation and estimation organized by the Center for Computational Statistics and Machine Learning at UCL. An abstract of his talk can be found below.

Fundamental to understanding how our world is represented in our brain is the ability to observe the collective activity of ensembles of neurons acting in concert while we associate sensory stimuli with subsequent motor actions. Some recent technological advances have greatly accelerated our ability to simultaneously record and stimulate these ensembles, thereby opening up the possibility to tremendously advance our understanding of the brain and its inner workings, and to provide real time neural control of assistive devices to people with severe disabilities through sophisticated brain-machine interfaces (BMIs). These remarkable advances, however, have outstripped progress in statistical signal processing theory and algorithms specifically tailored to: 1) analyze the massive amounts of neural and behavioral data collected; 2) explain many aspects of the natural information processing mechanisms in the nervous system; and 3) perform real-time neural signal processing within the resource-constrained environment of an implantable system for clinically viable BMI applications. 

In this talk, I will briefly describe some of our recent efforts towards achieving these goals. I will first discuss how some guiding principles from systems and computational neuroscience, machine learning, and compressive sensing literature may provide some useful insight into characterizing the dynamics of functional brain networks during task-specific behavior; and how changes in these dynamics, for example, during task learning or recovery from injury, can be monitored and quantified. Next, I will show how engineering a wireless, fully implantable neural interface system capable of extracting the hypothesized constituents of the neural code in a freely behaving subject can greatly help in this regard. I will conclude with a brief discussion on the potential of this framework to increase the reliability and effectiveness of brain-machine interfaces, and to develop better algorithms for statistical inference in complex networks in the face of uncertainty and nonstationarity.