AUTOMATIC SPIKE SORTING

Plexon, Inc., Dallas TX

Investigators

Abstract

The long term objective is to develop a totally automated procedure for detecting and separating extracellular neural action potentials (spikes) in real time. Such an accomplishment would have profound implications for the broad study of connectivity in the nervous system and for nervous system interface devices to aid certain physically impaired individuals. The more modest aims are to implement and evaluate a class of spike clustering algorithms and statistical quality measures of their performance, with a fairly robust off-line spike sorter product as an expected result. In Phase I we will build an off-line environment suitable for running a variety of algorithms against both a set of simulated data, comprised of real spikes and synthesized noise, and actual data streams. Selected procedures will be implemented for each of the six steps in the processing chain: detection, alignment, feature extraction, statistical description, clustering, and decision. Statistical models will be developed to estimate the accuracy of the sorting process & feed back quality measures to the user. The various algorithms will be rigorously tested against both simulated data and collections of real data contributed by several laboratories. Phase II will cover a much broader range of algorithms as well as on-line implementation. PROPOSED COMMERCIAL APPLICATIONS: The immediate application for this technology is to greatly simplify the task of setting up many channels of action potential (spike) sorting parameters for those neuroscientists using large scale extracellular recording as a research tool. This will be accomplished by enhancing Plexon's MAP on-line neural data acquisition system and as a software package to process data files acquired by certain other acquisition programs. The long term potential is the technology basis for any kind of permanent, spike-based neuroprosthetic devices.