10.5281/zenodo.4282448
https://zenodo.org/records/4282448
oai:zenodo.org:4282448
Oleg Lobachev
Oleg Lobachev
0000-0002-7193-6258
Leibniz-Fachhochschule
Alexander Pfaffenroth
Alexander Pfaffenroth
Hannover Medical School
Christoph Wrede
Christoph Wrede
Hannover Medical School
David Haberthür
David Haberthür
0000-0003-3388-9187
University of Bern
Ruslan Hlushchuk
Ruslan Hlushchuk
0000-0002-6722-8996
University of Bern
Thomas Salaets
Thomas Salaets
KU Leuven
Jaan Toelen
Jaan Toelen
KU Leuven
Evaluating registrations of serial sections with distortions of the ground truths. Supplemental data
Zenodo
2020
registration
ground truth
lung
2020-11-20
https://arxiv.org/abs/2011.11060
https://github.com/olegl/distort
10.1109/ACCESS.2021.3124341
10.5281/zenodo.4282447
https://zenodo.org/communities/ieee
Creative Commons Attribution 4.0 International
Evaluating Registrations of Serial Sections With Distortions of the Ground Truths
This is the supplemental data for our paper on how to benchmark registrations of serial sections with ground truths. The files are named as follows:
*_challenge.7z: local distortions and global rigid transformations applied, the input for the benchmark we used. Use this to test your rigid and non-rigid methods.
*_local-only.7z: only local distortions applied.
*_local-DIST.7z: the distortion maps for local distortions.
*_SURF-rigid.7z: local distortions and global rigid transformations applied, rigid transformations undone with SURF-based rigid-only method. Local distortions remain. Use this if your method does not cope well with large rigid transformations.
_*vis.7z: visualizations of distortions.
_rigid_ground.7z: the real rigid transformations used in the global phase.
*_ground.7z: the ground truth. All data fit each other, no distortions. Use this to compare your registration result to it.
There are three main modalities and one further, as a reference:
CT_*: µCT data, a rabbit lung, 600 images. (In ground truth, and local distortions, and global transformations we supply more images that went into the benchmark, 50 more from both beginning and end.)
EM_*: an EM serial block-face (SBF-SEM) data set of adult mouse lung, 1000 images. (EM ground truth is individually normalized, see paper.)
LS_*: a lung from the light sheet microscopy from a male 24 week-old rat, 300 images. (LS ground truth is individually normalized, too.)
REAL_*: a region from real serial sections from a rabbit lung, 2 images.
We also supply elastix parameter files.
A preprint has been uploaded to arXiv. The definite version is available from IEEE. The source code of the distorter is available from GitHub.
This work was supported by DFG grant MU 3118/8-1.
This work was partially supported by JST, PRESTO grant number JPMJPR2025, Japan.
This research was supported by a C2 grant from KU Leuven (C24/18/101) and a research grant from the Research Foundation – Flanders (FWO G0C4419N).
None of the funding bodies was involved in the design or execution of the study.