Single channel Tiffs in Vitessce

[lee-t]

I'm trying to get Merfish images from https://datadryad.org/stash/dataset/doi:10.5061/dryad.jm63xsjb2 which are two single-channel tiffs with 8 z-layers, one dapi stain and another for Na+/K+-ATPase. I would like to combine these images with either SpatialExperiment or anndata for downstream analysis but I'm having issues with these data formats; they tend to break on image data when using the supplied wrappers and I don't think there is any support for these images in vitessce anyways if I tried to write my own wrapper. Is there a way to import single channel tiffs into vitessce or viv? This way I can separate the cell-label, cell-boundary data, and molecule data and plot that spatially over the tiffs for the stain. Otherwise, I'll need to find a way to get this data into ome-tiff or ome-ngff so I can visualize the data in Vitessce.

[keller-mark]

We are working on this feature in the underlying image rendering library Viv hms-dbmi/viv#609 but currently only OME-TIFF and OME-NGFF (v3) are supported.

[lee-t]

Any suggestions then on how to go from single-channel tiffs into OME-TIFF and OME-NGFF?

[keller-mark]

To go from multiple single-channel tiffs to one OME-NGFF:

pip install tifffile numpy zarr numcodecs ome-zarr==0.2.1

import zarr
import numpy as np
from tifffile import imread
from ome_zarr import writer

c1 = imread("my_image_c1.tif")
c2 = imread("my_image_c2.tif")
my_image = np.stack((c1, c2)) # if using axes = "czyx" in write_image, then the numpy array axes should also be orderd czyx

z_root = zarr.open_group("my_image.zarr", mode = "w")

default_window = {
    "start": 0,
    "min": 0,
    "max": 65_535, # may need to change depending on the numpy dtype of the my_image array
    "end": 65_535 # may need to change depending on the numpy dtype of the my_image array
}

writer.write_image(
    image = my_image,
    group = z_root,
    axes = "czyx",
    omero = {
        "name": "My image",
        "version": "0.3",
        "rdefs": {},
        "channels": [
            {
                "label": f"Channel {i}", # may want to use a different name for each channel
                "color": "FFFFFF", # may want to use a different color for each channel
                "window": default_window
            } for i in range(my_image.shape[0])
        ]
    },
    chunks = (1, 1, 256, 256),
)

To go from multiple single-channel tiffs to one OME-TIFF:

You can use FIJI to merge multiple single-channel tiffs into a single multi-channel tiff.

• From the GUI, see Image -> Hyperstacks
• From a script, see t_to_z.js and corresponding Snakefile (lines 10 and 43)

Then use bftools (which comes with the bfconvert command line tool) to convert from tiff to OME-TIFF.

bfconvert -tilex 512 -tiley 512 -pyramid-resolutions 6 -pyramid-scale 2  -compression LZW my_input.tiff my_output.ome.tiff

To add channel names to OME-TIFF

If you want to add channel names, you will need to set them via OME-XML which you can get/set within an OME-TIFF using tiffcomment (which also comes with bftools)

The channel names come from the “Name” property of the Channel nodes in the OME-XML. See OME -> Image -> Pixels -> Channel.Name.

Then the bftools/tiffcomment command line tool can be used to get and set OME-XML metadata in an OME-TIFF.

tiffcomment my_image.ome.tiff > in.ome.xml
python add_channel_names.py -i in.ome.xml -o out.ome.xml
tiffcomment -set 'out.ome.xml' my_image.ome.tiff

In add_channel_names.py I used the beautifulsoup package to generate XML - it was before I knew about the ome-types package which has functions for writing OME-XML.

[keller-mark]

I recommend opening OME-TIFF files using FIJI and inspecting the OME-XML for debugging metadata such as channel names, etc.

[lee-t]

Thanks a bunch for that Mark! that gives us certainly some directions to follow up on. I also recently noticed we do not have scalefactors or tiling but if these tags are optional then I'm definitely going to give your code a try.

[cgohlke]

a way to get this data into ome-tiff

import tifffile
from scipy.ndimage import zoom

channels = ['dapi_stack.tif', 'membrane_stack.tif']

czyx_stack = tifffile.imread(channels)

levels = 4
tile = (256, 256)
compression = 'adobe_deflate'

with tifffile.TiffWriter('czyx_stack.ome.tiff', bigtiff=True) as tif:
    tif.write(
        czyx_stack,
        subifds=levels,
        tile=tile,
        compression=compression,
        metadata={'axes': 'czyx', 'Channel': {'Name': channels}},
    )
    for level in range(levels):
        czyx_stack = zoom(
            czyx_stack, (1, 1, 0.5, 0.5), order=1, prefilter=False
        )
        tif.write(
            czyx_stack,
            subfiletype=1,
            tile=tile,
            compression=compression,
        )

[lee-t]

Thank you for looking at the data! This proved easier than using FIJI ironically.

[lee-t]

Thanks to @cgohlke and @keller-mark I end up with this configuration for vitessce. As a follow up, is there a way to add in molecule or cell set data from another dataset?
If I configure a viewer in vitessce

vc = VitessceConfig(name='merfish', description='MERFISH measurements in the mouse ileum')
dataset = vc.add_dataset(name='ileum').add_object(AnnDataWrapper(gdata, 
    cell_set_obs=["leiden_final"], cell_set_obs_names=["Leiden"], spatial_centroid_obsm="spatial" , expression_matrix="X" ))
dataset2 = vc.add_dataset(name='jeff').add_object(
    OmeTiffWrapper(img_path="czyx_stack.ome.tiff", name='ileum'),)
spatial = vc.add_view(cm.SPATIAL, dataset=dataset)
tiff = vc.add_view(cm.SPATIAL, dataset=dataset2)

I get a vitessce instance with 2 coordinated spatial windows, one with the tiff and a corresponding plot with centroids and cell labels. Is there a way to overlay dataset onto dataset2 in vitessce? doesn't seem to work that way unless I'm missing something. Otherwise the next step I guess is to annotate the ome-tiff with cell labels or covert to ome-ngff while annotating labels such that the image data and cell set metadata are the same source dataset.

[lee-t]

Nevermind, I read that add_dataset() has an argument that accepts a list of objects. This will work:

vc = VitessceConfig(name='merfish', description='MERFISH measurements in the mouse ileum')
dataset = vc.add_dataset(name='ileum', objs= [AnnDataWrapper(gdata, cell_set_obs=["leiden_final"], cell_set_obs_names=["Leiden"], mappings_obsm=["X_umap", "X_pca"], mappings_obsm_names=["UMAP", "PCA"], spatial_centroid_obsm="spatial" , expression_matrix="X"), 
    OmeTiffWrapper(img_path="czyx_stack.ome.tiff", name='ileum' )])

[keller-mark]

add_object also returns the dataset instance so you can also chain the calls like

dataset = (vc.add_dataset(name='ileum')
  .add_object(AnnDataWrapper(gdata, cell_set_obs=["leiden_final"], cell_set_obs_names=["Leiden"], spatial_centroid_obsm="spatial" , expression_matrix="X" ))
  .add_object(OmeTiffWrapper(img_path="czyx_stack.ome.tiff", name='ileum'))
)