Rapid and Agile Multi-Photon Optical Imaging Over Large Neural Volumes

Johns Hopkins University, Baltimore MD

Investigators

Abstract

Project Summary/ Abstract Multi-photon, depth-sectioning microscopes are of paramount importance in capturing neural activity with cellular resolution. Despite their impressive image quality and robustness to scattering, diffraction limited, scanning multi- photon microscopes face a fundamental trade-off between the field of view (FOV) and imaging speed, which arises from the need to sequentially visit every voxel of the three-dimensional space. Furthermore, attempts to improve the speed by parallelizing multi-photon imaging through camera-based acquisition suffer from poor performance in scattering environments such as mammalian brain tissue. Notably, neurons themselves only occupy approximately 10% of the volume of interest and at any given time only a fraction of these neurons are active. Thus neural activity is spatio-temporally sparse and conventional point scanning technologies are highly inefficient as they spend a large amount of time and optical power passing through regions that lack relevant neural activity. The focus of this proposal is to develop a fast, agile, and rapidly adaptable multi-photon imaging technology that can leverage the spatio-temporal sparsity of the neural signals. First (Aim 1), we will develop an optical hardware system to enable rapid volumetric scanning with high-resolution random access for multi- photon imaging. Our approach will combine rapid Y-Z scanning of a temporally focused illumination line with patterning of the illumination line in the X-direction using a synchronized digital micromirror device. This hybrid scanning approach will allow highly parallel excitation and independent excitation of every voxel of the scan. We anticipate that the hardware system will be capable of scanning the full volume in < 5 ms. Second (Aim 2), we will develop adaptive compressed sensing (CS) algorithms that will allow us to use the hardware system developed in aim 1 to reconstruct an image of the volume at > 20 Hz. We will implement a patch-based, 4D algorithm (3D-space and time) for video rate neural recording that will leverage the joint spatio-temporal sparsity of neural activity to minimize the number of volume scans and laser energy required to construct an image of the volume. Third (Aim 3), we will validate the performance of our high speed microscope by imaging neural activity across the brain of live transgenic zebrafish (Danio rerio) Tg(elavl3:GCaMP6f)jf1. Ultimately, we anticipate that the technologies developed through this effort will provide a fast, agile, and versatile tool for optical recording deep inside live mammalian brain and other live animal tissues.