A Scalable and Modular Automated Pipeline for Stitching of Large Electron Microscopy Datasets

Abstract

Serial section Electron Microscopy can produce high throughput imaging of large biological specimen volumes. The high-resolution images are necessary to reconstruct dense neural wiring diagrams in the brain, so called connectomes. A high fidelity volume assembly is required to correctly reconstruct neural anatomy and synaptic connections. It involves seamless 2D stitching of the images within a serial section followed by 3D alignment of the stitched sections. The high throughput of ssEM necessitates 2D stitching to be done at the pace of imaging, which currently produces tens of terabytes per day. To achieve this, we present a modular volume assembly software pipeline ASAP (Assembly Stitching and Alignment Pipeline) that is scalable and parallelized to work with distributed systems. The pipeline is built on top of the Render [18] services used in the volume assembly of the brain of adult Drosophila melanogaster [2]. It achieves high throughput by operating on the meta-data and transformations of each image stored in a database, thus eliminating the need to render intermediate output. The modularity of ASAP allows for easy adaptation to new algorithms without significant changes to the workflow. The software pipeline includes a complete set of tools to do stitching, automated quality control, 3D section alignment, and rendering of the assembled volume to disk. We also implemented a workflow engine that executes the volume assembly workflow in an automated fashion triggered following the transfer of raw data. ASAP has been successfully utilized for continuous processing of several large-scale datasets of the mouse visual cortex and human brain samples including one cubic millimeter of mouse visual cortex [1, 25]. The pipeline also has multi-channel processing capabilities and can be applied to fluorescence and multi-modal datasets like array tomography.

Petascale pipeline for precise alignment of images from serial section electron microscopy

“…The connectomics team at the Allen Institute for Brain Sciences stitched and aligned the image stack to displacements within 20 µm using their ASAP framework (Mahalingam et al, 2021).…”

Plasmodium falciparumdisruption of pericyte angiopoietin-1 secretion contributes to barrier breakdown in a 3D brain microvessel model

“…The acellular center of the lumen was not imaged to save imaging time. Following acquisition, the positional metadata of the sections and tiles obtained was converted with SBEM-specific import scripts (https://git.embl.de/schorb/rendermodules-addons) for subsequent assembly, stitching and alignment of large EM datasets [67] . Aligned images were then binned to 22 by 22 by 250 nm 3 in xyz and segmentation was then performed using the software package Amira (Thermo Fischer Scientific).…”