pipeline_input

In this section is described the purpose and functionnement of the input.py script.

Assert folder integrity

The script begins with assessing the folder structure integrity. Using FISH FOLDER and LOCATION KEYWORD parameters it will detect the folders containing the image stacks and check all locations contains the same number of cycles (expected 1 ome.tiff per cycle). It then maps the path to every image stacks for each locations.

Assessing image stack structure

Next step is to assess image stacks structures and assert their similarity. In other words it will read images shapes and axes order. For example for an experiment with two imaging wavelength plus dapi, 3D stacks and 10 locations we could find the shape (10, 54, 2000, 2000, 3) with axes order (cycles,z,y,x,c).

The script first tries to open metadata from ome file structure. If you are building your own ome file structure to use this pipeline you can check this section explaining how to build ome metadata compatible with the pipeline. If metadata cannot be understood with pipeline logic it opens all image stacks and assess shape and axes order, again this need ome structure to function properly.

The structural information of image stacks and their path is stored in the Input table in the result_tables folder.

Ome files : Ome tiff files are meant to split an image stack in multiple files. Metadata can be described in all files, in an external files or only in first file. We exploit the later using tifffile package opening the first file (cycle0) we can access the whole image stack. If this logic fails the software will not be able to work properly.

Mapping cycles

Using MAP FILENAME input script opens the experiment file. Next using GENES_NAMES_KEY and CYCLE_KEY each cycle will be matched with the name of targed gene for each color.
During this process a few integrity checks are ran :

• All columns in GENES_NAMES_KEY are found in experiment file.
• No duplicate in gene names.
• Cycle number is equal in all image stacks.
• Cycle number found in image stacks match experiment file.

Additionaly, a warning can be raised : Some images registered in metadata were not found in folder. This means image stacks metadata describe more cycles that the number of file found in the location folder. In the case you manually deleted part of your cycle this can be expected behavior. In such a case the missing cycle will still be re-created with empty signal. Therefor they still need to be described in the experimental file.