The original pipeline for pre-processing, reconstructing, and template creation of the DAPI template.
The reconstruction and template creation was run twice. First with the allen reference atlas as reference without non-linear registration steps resulting in an initial DAPI template. The initial DAPI template was used as reference (instead of the Allen reference atlas) in the second run of reconstruction and template creation, resulting in the final DAPI template. The two runs of the pipeline were manually modified.
This pipeline is based on the original data structure of the slice images with a few sanity checks to make sure that all images are present. To run this pipeline with the data structure from the GIN-repository you need to remove these sanity checks, get the mouse ids from a different directory (like tif/), and get all the image paths in each individual tif folder.
The main script calls the pre_processing, co_registration, and template_creation with appropriate parameters and paths. In pre_processing the original tif image is down-sampled to 8.6x8.6 um and masked according to the biggest (except background) connected component after smoothing and threshold to remove artefact outside of the brain slice. After pre-processing the slices for each brain is reconstructed into 3-dimensional brain volumes. This reconstruction is carried out by iterative registrations (both between volumes and between slice-pairs) to a reference (Allen or DAPI from first iteration). Finally, each reconstructed brain volume is used in the creation of a population-based average in template_creation.
This pipeline uses the following non-standard packages (package name: version):
nibabel: 2.1.0
nipype: 0.13.1
numpy: 1.15.1
PIL: 5.1.0
scipy: 1.1.0
skimage: 0.14.0
sklearn: 0.19.2
A precompiled binary of Advanced Normalization Tools (ANTs) V.2.1.0 for linux/mac was used.