PORTLAND, Ore. -- An electrical engineer is determined to unify neural-prosthesis research by applying graphical state diagrams to bring together the disparate approaches taken by other experimental groups. Neural prosthetics convert brain activity into control signals that can drive electronics, but the algorithms that make the link have been unique to each implementation. Now, Lakshminarayan Srinivasan's state diagrams appear to offer a single method to conjure the intentions of a patient from the signals in his or her brain, then translate that into actuation of a prosthetic device.
"Neural-prosthetic devices, in general, are in their infancy," said Srinivasan, a recent graduate of the Massachusetts Institute of Technology's EE department. "My emphasis is to generate new perspectives on algorithm development that exploit both engineering theory and neuroscience insights, and study the science and engineering of prosthetic devices and movement control through human electrophysiology."
Neural prosthetic devices treat paralysis or amputation by implanting electrodes that monitor the remaining signals radiating from the patients brain whenever they intend to use the missing capability. Computer algorithms must not only sense these sometimes feeble signals, but, more important, must infer the intention of the patient to actuate the prosthetic device.
Unfortunately, until now researchers and prosthetic designers alike have been quick to begin crafting algorithms that are directly tied to the specific brain regions, recording modalities and underlying prosthetic application. However, by switching to a set of graphical state diagrams, like those used by EEs to design finite state machines, Srinivasan has been able to demonstrate an algorithmic approach that unifies these otherwise disparate efforts.
"The big challenge with each patient is to produce a device that works better than any other options," said Srinivasan. "Researchers with different backgrounds are attempting to attack these various design issues by leveraging their particular expertise."
Instead, Srinivasan applied his EE training in graphical state diagrams--circles that represent states connected with arrows that represent transitions among states--to make theoretical connections between sensed signals, indicated-intentions and prosthetic actuation. Whether the neural signals are sensed by implanted electrodes in an amputated limb, by intracranial electrode arrays, by electoencephalography (EEG) or by magnetic/optical imaging, they can all be integrated into Srinivasan's graphical state diagrams.
Instead of reinventing a new paradigm for each modality or brain region, Srinivasan proposes that multidisciplinary teams of engineers, neuroscientists and clinicians standardize on his graphical state diagrams. As a bonus, this approach should also enable next-generation future neural prosthetics to integrate multiple sensing modalities into a smarter prosthetic capable of more accurate actuation by virtue of checks-and-balances that cancel out false readings from individual sensors.
"I am also continuing to push the theoretical front to completely develop perspectives on optimizing neural prosthetic algorithms," said Srinivasan. "My next immediate concern is evaluating these algorithmic approaches through human studies."
To complete his training and acquire the necessary medical expertise to evaluate his own work, Srinivasan has entered medical school. He is currently a postdoctoral researcher at the Center for Nervous System Repair at Massachusetts General Hospital, as well as a medical student in the Harvard-MIT Division of Health Sciences and Technology (HST).
Srinivasan began working on his neural-prosthetic algorithms as a graduate student in MIT's Department of Electrical Engineering and Computer Science (EECS), where he graduated with his Ph.D. in 2006.