Large scale image processing with Nextflow

Biomedical image analysis is a rapidly expanding field and presents unique challenges for pipelines and workflows. Terabyte-scale microscopy data pushes the limits of existing workflow solutions, while also requiring the integration of diverse software solutions for block-based image processing tasks such as image stitching, registration, machine-learning-based segmentation, and spot detection. I will describe the EASI-FISH analysis pipeline for spatial transcriptomics, show some of the Nextflow DSL2 workflows we developed, and discuss the approaches we used to deal with large image data sets and polyglot tool sets (MATLAB, Java, Spark, Python, Dask).