The CosMx reader implements some pretty rigid expectations about image dimension order in input images here:
https://github.com/scverse/spatialdata-io/blob/main/src/spatialdata_io/readers/cosmx.py#L197
In the course of testing a pipeline for processing data for the SenNet and HuBMAP consortia, I've encountered data where dask_image.imread.imread() returns data with dimensions ordered like this:
>>> img = imread.imread('FOV0046.tiff')
>>> img
dask.array<_map_read_frame, shape=(5, 4256, 4256), dtype=uint16, chunksize=(1, 4256, 4256), chunktype=numpy.ndarray>The transposition of C and Y dimensions relative to the reader's expectations results in some issues visualizing data downstream. I understand that there's a new CosMx reader in development, and am wondering if it will have more flexible expectations about image dimension ordering, and if there's a timeline for the release of the new reader.
The CosMx reader implements some pretty rigid expectations about image dimension order in input images here:
https://github.com/scverse/spatialdata-io/blob/main/src/spatialdata_io/readers/cosmx.py#L197
In the course of testing a pipeline for processing data for the SenNet and HuBMAP consortia, I've encountered data where dask_image.imread.imread() returns data with dimensions ordered like this:
The transposition of C and Y dimensions relative to the reader's expectations results in some issues visualizing data downstream. I understand that there's a new CosMx reader in development, and am wondering if it will have more flexible expectations about image dimension ordering, and if there's a timeline for the release of the new reader.