An Ultra High-Density Virtual Array with Nonlinear Processing of Multimodal Neural Recordings

University Of California, San Diego, La Jolla CA

Investigators

Linked publications & trials

Abstract

An Ultra High-Density Virtual Array with Nonlinear Processing of Multimodal Neural Recordings A major goal of neuroscience is to record the activity of all neurons in an area of an intact brain and understand the relationship between neural activity and behavior. However, with current technologies, it is not feasible to have a direct and simultaneous access to every neuron in a three-dimensional brain area. Here we propose a novel approach, combining an innovative signal processing method with optical and electrical recording technologies to `virtually' record from all neurons in a three dimensional volume. If successful, this approach will allow us to substantially increase the number of recorded neurons without the need for direct optical or electrical access to each neuron. The proposed Virtual Array technology has the potential to dramatically increase the number of simultaneously recorded neurons in an intact brain relatively non- invasively. The common approaches include high-density electrophysiological probes, which are highly invasive and also limited in the density of recording, and fast-scanning optical techniques that have limited temporal resolution. As an alternative approach, we propose to develop a framework to computationally increase the number of recorded neurons out of recording data from simultaneous electrophysiology and imaging. The computational framework will be developed from a dataset in which micro- electrocorticogram (µECoG) are recorded simultaneously while the activities of the underlying neurons is recorded with two-photon calcium imaging at multiple cortical depths. We will virtually reconstruct this single-cell activity by solving appropriate optimization problem involving forward models for µECoG recordings and calcium signals. This optimization problem will be solved using alternating convex algorithms.