Research

Optimized detection and decoding of memory-related patterns of neural activity
Particular, temporally-limited patterns of neural firing appear to be one of the mechanisms which enable information to be reliably stored and transmitted between different brain regions and to/from the periphery. Most engineering study of these patterns has focused on neuro-prosthetic interfaces. For example, neural decoding to control prosthetic limbs (e.g., here, here, and here), or spatio-temporally structured stimulation to provide artificial sensory representations. What are the patterns of neural activity that enable emotion, decision making, and memory? Recent results have suggested that neural firing during hippocampal sharp wave ripple oscillations may specifically underly particular forms of memory storage and recall (here and here and here). What are the limits of detecting, decoding, and controllably manipulating these patterns of activity in real-time?

Open source systems for real-time decoding and manipulation of neural circuits in vivo (ARTE project)
In the last decade, it has become clear to neuroscientists, engineers and clinicians that there are critical scientific questions and significant potential therapeutic benefits from manipulating neural activity based on on-going neural recordings. This has become even more important as our toolbox for controllably activating or inactivating neural circuits on fine time scales expands beyond electrical stimulation. Unfortunately, studies of this type have had to spend significant amounts of time and resources repeatedly building custom systems to integrate neural recording, real-time information extraction and control signal generation. We are building a low-cost, high performance, open source hardware and software platform which will allow investigators to focus their development efforts on the particular algorithmic challenges for a given experiment.
 
Embedded systems for extended, intelligent perturbation of cognitive neural circuits (information lesions)
For almost a century, neuroscientists have learned about the brain by studying the cognitive and behavioral deficits which result from lesions of particular brain regions. Though these experiments have evolved from the observation of unintentionally injured individuals to the controllable silencing of genetically-targeted populations of cells, the basic idea remains the same - remove some neurons from the circuit and see what happens. We know, however, that the structure and context of the activity of a neural ensemble appears to radically change how it affects downstream neurons. Thus, we seek to enable the next generation of lesion studies, where neur
al ensembles are selectively inhibited based on the information they are currently representing. Toward this end, we are developing systems which can decode ensemble neural activity and selectively manipulate the circuit in an unsupervised manner with limited power for extended periods of time (days or months).