Stephan Preibisch: “Reconstructing large-scale lightsheet microscopy acquisitions”

Stephan Preibisch, Team Lead, Computational Methods Development at HHMI's Janelia Research Campus. Abstract: Light-sheet microscopy is an emerging technology that allows the acquisition of developing samples like Drosophila or mouse embryos in their entirety, and recently also enables to image very large fixed samples such as adult mouse brains at single-cell resolution or even Drosophila larval brains with super-resolution. To cover entire samples, sets of tiled three-dimensional images can be acquired from different physical orientations while illuminating the sample from different directions. These image acquisitions amount to very large datasets in the terabyte range that require complex image reconstruction including image registration, correction of optical distortions arising from light refraction inside the sample, and deconvolution in order to create seamless datasets for downstream analysis (Fig. 1). We will present concepts and software that efficiently handles and interactively reconstructs such large multi-tile, multi-view acquisitions. These solutions automatically and efficiently align all image tiles, optionally compensate aberrations inside the sample using non-rigid image deformation, and fuse datasets using averaging strategies or deconvolution. Additionally, image resolution can be estimated in all parts of the reconstructed sample to confirm uniform, high-resolution acquisitions. The software is implemented as open-source based on ImgLib2, uses BigDataViewer and Spark, and large parts are available as Fiji plugins making high-resolution light-sheet microscopy acquisitions amenable to biological studies. We will additionally discuss new developments for data storage, handling of sheared volumes and visualization that will in the future to scale to up to peta-byte sized image acquisition. More information about Janelia’s Optical Interest Group can be found here: https://www.janelia.org/OIG