Faculty Seminar by Nicholas Pegard from UC Berkeley.
Title : Computational Optics Beyond Imaging: New Instrumentation to
Monitor and Manipulate Neural Activity with Light.
Future progress in neuroscience research and medicine requires
high-performance instruments that can monitor and manipulate brain
activity on demand, in 3D, and at high speeds. With the recent
development of optogenetics and optical reporters of neural activity, it
is theoretically possible to measure, trigger or inhibit activity in
individual neurons with light. However, current optical instrumentation
is unfit to address these new challenges since it relies on recording
then processing large 3D images, introducing inefficiencies and limiting
the number of neurons that can be simultaneously addressed.
In this presentation, we propose new optical instrumentation and
algorithms that are jointly designed and optimized to perform
brain-machine interfacing tasks without the need for image
reconstruction and processing. We first show how functional fluorescence
signals (e.g. with GCaMP) can be quantified simultaneously in many
neurons and in 3D by directly processing a single raw data frame
acquired with a light-field microscope. Specifically avoiding image
reconstruction improves speed and better preserves relevant quantitative
information in the presence of strong optical aberrations.
We then consider the reverse problem of sculpting light in 3D to
photostimulate custom neuron ensembles on demand. For this, we have
developed a new multiphoton illumination technique called 3D-SHOT, which
combines 3D computer generated holography and temporal focusing.
Experiments in awake mice demonstrate precise targeting capabilities
with high temporal precision and single neuron spatial resolution in
large volumes, outperforming other image-forming strategies.
Finally, we present new algorithms to compute better holograms by
solving an optimization problem with an explicit cost function. For
applications in neuroscience, we show how performance can be further
improved by tailoring the cost function to account for known biological
properties of brain tissue and to optimize for the desired task outcome.