Skip to main content

Highlight

Whole-brain, 3D imaging of neural networks in vivo

Achievement/Results

NSF-funded researchers at Stanford University have developed a new way to look at 3D circuits in the brains of behaving animals. By combining novel computational imaging techniques with indicators of neural activity, they can record a 3D volume of neurons at every camera exposure, and then extract information about what thousands of different neurons are doing in the volume over time. By applying this method to tiny, optically transparent larval zebrafish, it is now possible to simultaneously look at the activity of neuronal circuits across the whole brain volume of an awake, behaving vertebrate under a microscope.

The project is a result of interdisciplinary work funded by the NSF Integrated Graduate Education and Research Traineeship (IGERT) program, which has allowed Stanford Neurosciences Ph.D. candidate and IGERT Trainee Logan Grosenick to apply light field microscopy 1 —a technique pioneered by IGERT co-mentor Marc Levoy (Professor of Computer Science and of Electrical Engineering)—to all-optical interrogation of brain circuitry in the lab of co-mentor Karl Deisseroth (Professor of Psychiatry and of Bioengineering) as part of an IGERT bridging research project. The project uses techniques from computational photography to render camera images into volumes, allowing scanless 3D imaging of large tissue volumes at video rates. Recently developed machine learning algorithms are then applied to the time-varying volumes to first extract neural activity, and then to relate this activity to what the fish sees and does.

By synchronously capturing how neural activity propagates through the complicated 3D network of neurons in an animal’s brain, and relating this cell- and circuit-level activity to the animal’s perception and behavior, this technique will give researchers an unprecedented look at how brains process information and generate actions. Many activities important to survival, such as eating, sleeping, learning, or just tracking an object visually, require the coordinated effort of 3D circuits that interconnect across the brain. Because light field neuroimaging can be used to rapidly image circuits in large tissue volumes, it will yield insight into how such fundamental processes are shaped by the coordinated activity of neurons across the brain, and how failures of neural circuitry may lead to pathological conditions related to disease.

Address Goals

This project advances the frontiers of both neuroscience and computational optics by synthesizing new technologies from the two fields in order to yield fast, completely synchronous 3D imaging of neuronal activity inside a living animal. For neuroscience, this means the opportunity to look at whole brains with single-neuron resolution at video rates in vivo. For computational optics it presents new opportunities and challenges surrounding diffraction limited optics and biological samples. The marriage of the two technologies also results in incredibly large, information-rich datasets that have already inspired new developments and applications of machine learning algorithms for 3D image data [2,3].

This IGERT project has fostered learning directly, as the Stanford Center for Mind, Brain, and Computation IGERT program requires all trainees to design and complete a course of study—-in this case Grosenick completed classes in optics, machine learning, optimization, and optogenetics, all relevant to the bridging research project. There are also more general consequences of the project that relate to making fast 3D imaging accessible and easy for a broader audience. Because light field microscopy is relatively inexpensive and it is easy to construct a light field microscope (the complicated details have been pushed to freely available software), adoption of the technology is easy, and should allow it to be used broadly in both scientific and educational settings. This is in contrast to laser-based scanning methods commonly used for imaging neural activity, which are often both technically challenging and prohibitively expensive.