Optical Oscilloscope: Real-time, High-throughput, Volumetric Voltage Imaging

Neuroscientists lack tools to monitor the voltage of thousands of neurons at cellular resolution and in real time. This is a problem because voltage conveys the context and content of neuronal network dynamics, functional connectivity, and excitability which continually evolve. Unlike spiking, which can be monitored by calcium imaging and electrode arrays, no existing methods track subthreshold voltage fluctuations at network scale. Multiphoton scanning modalities, typically used to mitigate light scattering by brain tissue, collect fluorescence too slowly to resolve small, fast voltage signals in >100 neurons at a time. Our goal is to develop network-scale, volumetric voltage imaging capability based on light-field microscopy (LFM). LFM enables scanless, light-efficient volume acquisition, but its degradation by light scattering and computational complexity hinder its application to network-scale voltage imaging. We will overcome these limitations through a new generation of optics-aware, computationally efficient deep neural networks (DNNs) that extract neuronal voltage signals from scattering-corrupted light fields. We will train the DNNs with both one-photon light fields and scattering-robust two-photon scanned volumes to leverage the advantages of these complementary modalities. We will implement the DNNs in a field-programmable gate array for real-time kilohertz voltage readout to enable closed-loop experiments. Our Optical Oscilloscope addresses the unmet need for low-latency, cellular-resolution, network-scale voltage readout.  These new capabilities will catalyse discovery of how neural networks process, learn, and remember information, and how these functions are compromised in diseased and injured brain tissues.  It will also form the basis for a new generation of electro-optical brain machine interfaces.