Get Your Brain Together Hackathons

Back to the Tutorials List

Large Scale Registration

Instructors

Tutorial Description

Lightsheet microscopy methods for mouse brains result in volumes that are not feasible to register at full resolution on a standard single computer. As we move to register rats, non-human primates and whole human brains, a scalable registration approach is required that is not limited by the memory capacity of modern computational nodes and the computational capability of a single system.

In this tutorial, we will cover principles for large scale registration that combine the OME-Zarr NGFF, Xarray, and Dask, a distributed computing library for Python, to register neuroimages in a way that can limit memory consumption, distribute computation, and perform a coarse-to-fine registration that progressively aligns structures across multiple scales.

Duration: 0.5 hours.

Learning Outcomes

  1. Outcome 1. Understand the need for multiscale OME-Zarr spatial metadata and how to generate this metadata.
  2. Outcome 2. Learn how to distribute registration to limit memory consumption and improve performance.

Approach and Materials

Background and References