Project Summary: The goal of this project is to create two prototypes of a novel live spike sorting system which can be used by investigators to spike sort streams of neural data recorded by multi-channel, high channel and ultra-high channel probes. In most in-vivo extracellular recording conditions, an electrode can pick up neural spikes from several nearby neurons resulting in so-called “multi-unit” activity in the recording trace. Spike sorting algorithms are then used to separate this multi-unit activity into several sets of “single-unit” activities, each of which represents the action potential firing pattern of a single neuron. This sorting process is typically a computationally intensive process and is growing into a critical technology gap with the advent of multi and high channel count hardware. Live spike sorting of a complete set of multichannel data has been challenging if not impossible. On the other hand, there is a demand for live spike sorting during an experiment, especially by those investigators who record from functionally heterogenous brain areas such as, for example, all cortical regions. If an investigator had the ability to review live single cell data, he/she could determine the quality of the data and adjust the electrode position or decide on next experimental steps based on the incoming results. We recently developed the GEMsort algorithm, which, compared to existing spike sorting algorithms, was designed to sort neural spikes from multichannel probes with immediate sorting outcomes. These algorithms provide powerful, accurate yet computationally inexpensive spike sorting due to a different mathematical approach. As a result, these algorithms can spike sort complete streams of complex data, including data recorded with high channel and ultra-high channel electrodes virtually in real time. In this proposal, we will develop two tabletop-sized systems based on Field-Programmable Gate Array (FPGA) technology for laboratory use. These systems will be based on the GEMsort algorithm and add live spike sorting capabilities to an investigator's existing recording setup.

Narrative: Neural spike sorting is a computational technique to calculate the firing pattern of single neurons from multi-unit activity measured by a recording channel of a neural probe. We developed a novel algorithm, GEMsort, which requires significantly less computational power, facilitating live spike sorting during in-vivo recordings, even for high channel recording setups. We propose to build two systems based on this algorithm that can be used to review spike sorted data in even hundreds of channels live during an experiment, which will allow an investigator to assess the quality of the incoming data such that the experiment can be modified on the fly as needed.