the input for each step (i.e., task within the pipeline) should have one input param, i.e, the image to be processed.
The image can be the full image or a tile - this basically does not really matter.
Additionally, there should be a globally accessible "meta data store" (i.e., dictionary) where all information about the tiles can be stored (i.e., whether it is tumor or non-tumor, ...)
This store can be accessed dynamically from all steps
the input for each step (i.e., task within the pipeline) should have one input param, i.e, the image to be processed. The image can be the full image or a tile - this basically does not really matter.
Additionally, there should be a globally accessible "meta data store" (i.e., dictionary) where all information about the tiles can be stored (i.e., whether it is tumor or non-tumor, ...)
This store can be accessed dynamically from all steps