MIT, Singleton Auditorium
A major hurdle to understanding how neural activity leads to behavior is determining which spikes originated from which neurons, a problem known to electrophysiologists as spike sorting. This problem becomes easier when sensors of neural spikes are small and more numerous than neurons in their recordable volume. This fact motivated the recent design of “ultra-dense” multi-electrode arrays (MEA) with many close-packed sensors (e.g. 10um x 10um recording sites spaced 1.5um apart with 256 to a shank).
Here Allen presents results from MEA recordings he conducted in the visual cortex of lightly-anesthetized and awake mice in concert with co-localized patch clamp recordings. As expected, signals recorded on sensors of the MEA contained mixtures of spikes from many neurons, and patch clamp recording revealed unambiguously the spikes of a single neuron. The latter could then be used as “ground truth” for spike sorting evaluation, and the patched neuron’s spike times could be used as a trigger to reconstruct its spike field across the MEA, revealing its extracellular signature in unprecedented detail.
Allen will discuss how properties of spikes revealed through these recordings may vary with brain state, and will touch on the methodology behind the data. This methodology includes the development of a real-time feedback system for determining inter-probe distance— a difficult problem when precision on the order of tens of microns in living tissue is required— that may as a by-product prove useful for studying basic electrical properties of brain tissue. Finally, building off of previous computational modeling work, Allen will present promising results demonstrating the use of a blind source separation technique to spike sort these recordings, and will motivate the adoption of this dataset, or “library of cells”, to become the new gold standard for spike sorting evaluation.
Host/Chair: Edward Boyden
Ki GoosensNancy Kopell