Optimising and generalising image processing workflows for cleared mouse brains imaged using light sheet microscopy

Session

Poster Session

Authors

Olivier Burri (1), Lorenzo Talà (1), Nicolas Chiaruttini (1), Rémy Dornier (1), Axel Bisi (2), Lana Maria Smith (2), Sylvain Crochet (2), Carl C.H. Petersen (2), Arne Seitz (1)

Affiliations

1. EPFL - BioImaging and Optics Platform
2. EPFL - Laboratory of Sensory Processing

Keywords

lightsheet holder clearing

brains stitching registration cluster hpc snakemake docker

Abstract text

The scanning of cleared mouse brains using light sheet microscopy has become increasingly routine and continues to evolve (iDISCO1, fDISCO2). The acquisition of whole-brain 3D images generally requires delicate sample mounting steps, followed by tiled acquisitions. Before being able to extract quantitative information from images to answer biological questions, a general "preprocessing" workflow involving tile registration, channel alignment, tile fusion and brain reorientation is required. Moreover, the resulting fused dataset must usually be in a particular format (stack, individual planes per image) and orientation depending on the tools used downstream (BrainGlobe’s Brainreg3, Elastix,…) 

Luckily, there are softwares available to tackle such problems (BigStitcher4, TeraStitcher5, Stitchy), however they were originally built around the idea that this process is "low throughput", and contain several steps that could benefit from optimization.

To tackle the growing need for acquiring and preprocessing a large number of samples, several improvements have been worked on.

1. An improved holder aimed at long-term storage of iDISCO-cleared brains, allowing a fast switch from storage to imaging with minimal sample damage during initial mounting.

2. An optimised "Quick CZI reader" in BioFormats, which avoids the need for a data duplication step while performing preprocessing.

3. A brain stitching and fusion workflow that uses BigStitcher through SnakeMake and Docker to enable deployment on either a local workstation or cluster-type architectures.

4. A direct way to output various data formats depending on the downstream processing. 

The various elements are implemented individually, allowing them to be transferable to other workflows. We combined them through scripts inside a Docker container. This will allow us to easily customise and package similar workflows suited to the needs of different users in the cleared lightsheet community.

References

1.    Renier, N. et al. iDISCO: A Simple, Rapid Method to Immunolabel Large Tissue Samples for Volume Imaging. Cell 159, 896–910 (2014).

2.    Qi, Y. et al. FDISCO: Advanced solvent-based clearing method for imaging whole organs. Sci. Adv. 5, eaau8355 (2019).

3.    Tyson, A. et al. Accurate determination of marker location within whole-brain microscopy images. Scientific Reports vol. 12 https://doi.org/10.1038/s41598-021-04676-9 (2022).

4.    Hörl, D. et al. BigStitcher: reconstructing high-resolution image datasets of cleared and expanded samples. Nat. Methods 16, 870–874 (2019).

5.    Bria, A. & Iannello, G. TeraStitcher - A tool for fast automatic 3D-stitching of teravoxel-sized microscopy images. BMC Bioinformatics 13, 316 (2012).