adjust for crc2 dataset #204
Conversation
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #204 +/- ##
==========================================
+ Coverage 35.91% 45.50% +9.58%
==========================================
Files 19 22 +3
Lines 1715 2147 +432
==========================================
+ Hits 616 977 +361
- Misses 1099 1170 +71
|
| load_image( | ||
| path=path / VisiumHDKeys.IMAGE_HIRES_FILE, | ||
| path=( |
There was a problem hiding this comment.
please make this check when constructing the hires_image_path (same for the other images) and add the path in a intermediate variable.
Something like:
root_level_spatial_path = path / VisiumHDKeys.SPATIAL
# if doens't exist
root_level_spatial_path = path / libary_id / VisiumHDKeys.SPATIAL
# if still doens't exist:
# raise exceptionAnd then please remove spatial/ from VisiumHDKeys.IMAGE_HIRES_FILE, ...
In this way the code will be more robust. At the moment spatial appears as a keyword in line 299, and it's hidden in VisiumHDKeys.IMAGE_HIRES_FILE otherwise.
| path=( | ||
| hires_image_path | ||
| if hires_image_path.exists() | ||
| else path / f"{filename_prefix}spatial" / VisiumHDKeys.IMAGE_HIRES_FILE |
There was a problem hiding this comment.
In general let's try to have dataset specific keywords, like spatial outside VisiumHDKeys otherwise typos could lead to subtle problems.
| fullres_image_filename = fullres_image_filenames[0] | ||
| fullres_image_file = path_fullres / fullres_image_filename | ||
| fullres_image_paths = [path_fullres / image_filename for image_filename in fullres_image_filenames] | ||
| elif list((path_fullres := (path / f"{filename_prefix}tissue_image")).parent.glob(f"{path_fullres.name}.*")): |
There was a problem hiding this comment.
Same comment here as below, please use dataset specific keywords only in VisiumHDKeys.
There was a problem hiding this comment.
For this one I did not change VisiumHDkeys. The thing here is that these images live within the top level directory where in the current code a microscope_image directory was assumed.
There was a problem hiding this comment.
In other words this code already follows sort of your comment of first try this, then that and otherwise raise exception.
There was a problem hiding this comment.
From what I can see with the mouse intestine dataset, the actual fullres image is in the top directory as well. Is the microscope_image directory valid?
|
Thanks for the PR, I have added some comments. Please after the edits, please verify that:
|
|
Nice! After verifying the code on the crc2 datasrt, it is good to merge! |
I added one part to make it more robust with shapes, it can now deal with any level of nestedness as long as binned_outputs is somewhere. |
When reading in the data from https://www.10xgenomics.com/products/visium-hd-spatial-gene-expression/dataset-human-crc a few problems were encountered. First, the
lowres,hiresetc images were located in a directory with as root level thedataset_idprefix followed by spatial which then contained thespatialdirectory. Furthermore the naming of the fullres image was different. This PR accounts for that.