High-content behavioral phenotyping of model organisms in three dimensions with a gigapixel microscope

Significance: High-throughput optical microscopy is currently transforming the research fields of genetics, drug discovery and neuroscience. Large-scale optical assays now routinely use thousands of high-resolution images to offer critical insights into the human body, our brain, and the diseases that affect us. Today’s optical microscopes and their associated image processing software, however, remain far from ideal. Current microscopes 1) cannot form images with cellular-scale resolution over an area larger than a few square centimeters, and 2) cannot readily obtain 3D measurements at high speed. This fundamentally limits experimental capabilities, most notably within model organism research. For example, current methods must constrain or sedate organisms like the zebrafish and Drosophila for in vivo observation at cellular-scale resolution in 3D, which naturally alters their behavior and leads to slow and cumbersome experiments. Proposal: Ramona Optics aims to create a new micro-camera array microscope platform (MCAM-3D) for parallelized high-content 3D imaging of unconstrained model organisms. Two unique features of its novel arrayed design will be integrated within a single device to record 3D behavior and morphology: 1) real-time stereoscopic 3D depth at video rates, and 2) the ability to image both specimen ventral (top) and lateral (side) morphology using a new “periscopic” well plate that is composed of an array of tilted mirrors. Jointly developed software for automated 3D image analysis will be developed to improve the rigor, efficiency, and translatability of current model organism research and to produce new discoveries. SA1 – MCAM-3D hardware and software for large-area, high-speed video: We will develop a 9x6 micro- camera array (770 megapixels total) capable of rapidly re-focusing to one of 3 imaging configurations (2D video, 3D depth, lateral view) and multi-gigapixel-per-second capture. Software will generate 3D height maps in near-real time (<0.1 sec/frame) and execute per-organism tracking and pose estimation. SA2: Lateral perspective imaging with the MCAM-3D: A custom periscope well plate will re-direct light to simultaneously provide 54 unique lateral viewpoints at 6 µm two-point resolution within each well. Software will automatically register images and execute organism tracking and pose estimation. SA3: Annotation software for 3D behavioral and morphological analysis: We will create open-source software for automated detection and tracking of user-defined features of interest. Machine learning software will form per-organism feature “dictionaries” for statistical analysis with clustering software. SA4: Experimental demonstrations of the MCAM-3D: We will complete a drug screen assay with the McCarroll Lab (sub-award, UCSF) to measure 3D feeding behavior, and a developmental assay with the Bagnat Lab (sub-award, Duke) examining lateral morphology, to solidify a product-ready MCAM-3D.