Open Call Project, 2025 Alignimal
Scenario
Researcher
Sanja Jasek, Centre for Organismal Studies
Initial problem
The researcher works with very large volume electron microscopy datasets. Small organisms are sliced into sections, which are scanned separately at extremely high resolution. The raw data then consist of hundreds up to thousands of sections, each of which is usually divided into several tiles. Due to the soft material being cut, several deformations are inevitably introduced during this process, for example “wrinkles” due to folding of the thin sections, local deformations due to stress, and rotation caused by handling each slice individually.
In order to reconstruct a three-dimensional model of the whole organism, the individual tiles have to be stitched together into sections (slices), which then have to be aligned with each other.
The open-source Python package feabas is designed for exactly this kind of workflow. However, at the time this project started, it had several shortcomings. In some cases, individual sections consisted of only one tile instead of many. Feabas did not support this case.
Outcome
What we did
- Helped enhance feabas such that single-tile stacks and mixed-tile stacks can be processed
- Provided bash scripts for mounting SDS on BwUniCluster3.0 correctly for jobs running in parallel
- Provided detailed instructions for setting up feabas workflows on BwUniCluster3.0
- Tested alignment for exemplary datasets. Corresponding visualizations can be found below.