Merging individual tif channels

[Justin-Krieg]

Hello, I've been using VALIS for a little while and works very well! However I have a set of files that are individual tiff files for each channel, ie.
R1-DAPI.tif
R1-IBA-1.tif
R2-DAPI.tif

My initial thoughts was use bfconvert to merge these into a pyramidal multichannel tif, and whilst the input ome.tif files open in Qupath, and VALIS runs these files through fine, the resulting output crashes Qupath and is unable to load, which might be something worth investigating further. I wasn't sure if VALIS had anything built in that would avoid using bfconvert

Cheers,

Justin

[cdgatenbee]

Hi @kriegy98,

Apologies for taking so long to get back to you, just had a few very time sensitive things I had to take care of. Anyhow, I might suggest a more direct way to do this would be to use valis to register all of the DAPI rounds, apply those transformations to the associated slides, and merge the channels into a single ome.tiff. Assuming all the images are in the same directory, that would look something like this (I added an extra round and made up some markers for illustrative purposes):

img_list = ["R1-DAPI.tif",
            "R1-IBA-1.tif",
            "R2-DAPI.tif",
            "R2-CD3.tif",
            "R3-DAPI.tif",
            "R3-CD8.tif"]


# Get DAPI images, ordered by staining round
dapi_imgs = sorted([f for f in img_list if re.search("dapi", f.lower()) is not None])

# Register dapi images (need full path) by passing the list of their filenames to `img_list`, which overrides `src_dir`
full_dapi_paths = [os.path.join(src_dir, f) for f in dapi_imgs]
registrar = registration.Valis(src_dir=src_dir, dst_dir=dst_dir, img_list=full_dapi_paths)
rigid_registrar, non_rigid_registrar, error_df = registrar.register()

# Create dictionary that maps each DAPI image to the other images from the same round (based on filename)
rounds_dict = {f:[g for g in img_list if f.split("-")[0] == g.split("-")[0] and g != f ] for f in dapi_imgs}

# Apply transforms to other images from the same round and merge into single slide
merged_slide = None
marker_names = []
for dapi_f in dapi_imgs:
    # Get DAPI slide
    dapi_slide = registrar.get_slide(dapi_f)
    warped_dapi = dapi_slide.warp_slide(level=0)
    marker_names.append(dapi_slide.name)
    if merged_slide is None:
        merged_slide = warped_dapi
    else:
        #Note: comment out this `else` block if you only want the 1st dapi channel
        merged_slide = merged_slide.bandjoin(warped_dapi)
    other_imgs_in_round = rounds_dict[dapi_f]
    for other_f in other_imgs_in_round:
        # Warp other slides from the same round as the DAPI slide
        full_path_to_other_f = os.path.join(src_dir, other_f)
        other_name = valtils.get_name(other_f)
        marker_names.append(other_name) # Note: you could use some other approach to convert the filename to the marker name
        
        warped_other = dapi_slide.warp_slide(level=0, src_f=full_path_to_other_f)
        merged_slide = merged_slide.bandjoin(warped_other)

# Create ome-xml
ref_slide = registrar.get_ref_slide()
px_phys_size = ref_slide.reader.metadata.pixel_physical_size_xyu
bf_dtype = slide_io.vips2bf_dtype(merged_slide.format)
out_xyczt = slide_io.get_shape_xyzct((merged_slide.width, merged_slide.height), merged_slide.bands)

ome_xml_obj = slide_io.create_ome_xml(out_xyczt, bf_dtype, is_rgb=False,
                                        pixel_physical_size_xyu=px_phys_size,
                                        channel_names=marker_names,
                                        colormap=None)
ome_xml = ome_xml_obj.to_xml()
# Save merged slide to `dst_f` with ome-xml
slide_io.save_ome_tiff(merged_slide, dst_f=dst_f, ome_xml=ome_xml)

The above approach will avoid the need to use bfconvert, but it's hard to know if it will prevent QuPath from crashing. If you try this out, please let me know how it goes and if you run into any issues.

Best,
-Chandler

[Justin-Krieg]

Hi Chandler that worked perfectly thank you!

For any python novice's trying to replicate this, i had to use this at the start:

from valis import registration, valtils, slide_io
import re, os
src_dir = "/var/tmp/Combine"
dst_dir = "/var/tmp/Output" 
dst_f = "/var/tmp/Output/merged_slides.ome.tiff" 

I did notice this warning pop up which i believe is a seperate issue:
/usr/local/src/valis/slide_io.py:1587: FutureWarning: Unlike other reduction functions (e.g. skew, kurtosis), the default behavior of mode typically preserves the axis it acts along. In SciPy 1.11.0, this behavior will change: the default value of keepdims will become False, the axis over which the statistic is taken will be eliminated, and the value None will no longer be accepted. Set keepdims to True or False to avoid this warning.
Once again thank you,

Justin

[sonechkina]

Hello Chandler,

I'm facing the exact same problem as Justin: I have two cycles of 4 channels each in .tiff format and I want to merge them into a multichannel pyramidal image in .ome.tiff.
Everything worked smoothly, but when I open the composite image on HALO all the channels of each cycle result as the DAPI of the same cycle.
In the following link you'll find the single channels and the merged result, and also a copy of the whole code that I used to generate the said image.
https://drive.google.com/file/d/1xGjdFuUvuO2NRzmJO2yERO_kwdb3JYla/view?usp=sharing

Thanks for your support!

Best,
Maria

from valis import registration, valtils, slide_io
import re, os


src_dir = '/usr/local/src/registration__valis/singoli_canali/'
dst_dir = '/usr/local/src/registration__valis/registration_output/'
dst_f = '/usr/local/src/registration__valis/output_images/merged_slides.ome.tiff' 


img_list = ["1_DAPI.tif",
            "1_CD1c-Cy3.tif",
            "1_CD14-AF488.tif",
            "1_FCeRIa-AF647.tif",
            "2_DAPI.tif",
            "2_CD5-AF647.tif",
            "2_CD16-Cy3.tif",
            "2_HLADR-AF488.tif"]


# Get DAPI images, ordered by staining round
dapi_imgs = sorted([f for f in img_list if re.search("dapi", f.lower()) is not None])

# Register dapi images (need full path) by passing the list of their filenames to `img_list`, which overrides `src_dir`
full_dapi_paths = [os.path.join(src_dir, f) for f in dapi_imgs]
registrar = registration.Valis(src_dir=src_dir, dst_dir=dst_dir, img_list=full_dapi_paths)
rigid_registrar, non_rigid_registrar, error_df = registrar.register()

# Create dictionary that maps each DAPI image to the other images from the same round (based on filename)
rounds_dict = {f:[g for g in img_list if f.split("_")[0] == g.split("_")[0] and g != f ] for f in dapi_imgs}

# Apply transforms to other images from the same round and merge into single slide
merged_slide = None
marker_names = []
for dapi_f in dapi_imgs:
    # Get DAPI slide
    dapi_slide = registrar.get_slide(dapi_f)

    warped_dapi = dapi_slide.warp_slide(level=0)

    marker_names.append(dapi_slide.name)
    if merged_slide is None:
        merged_slide = warped_dapi
    else:
        #Note: comment out this `else` block if you only want the 1st dapi channel
        merged_slide = merged_slide.bandjoin(warped_dapi)
    
    other_imgs_in_round = rounds_dict[dapi_f]
    
    for other_f in other_imgs_in_round:
        # Warp other slides from the same round as the DAPI slide
        full_path_to_other_f = os.path.join(src_dir, other_f)
        other_name = valtils.get_name(other_f)
        marker_names.append(other_name) # Note: you could use some other approach to convert the filename to the marker name
        
        warped_other = dapi_slide.warp_slide(level=0, src_f=full_path_to_other_f)
        merged_slide = merged_slide.bandjoin(warped_other)

# Create ome-xml
ref_slide = registrar.get_ref_slide()
px_phys_size = ref_slide.reader.metadata.pixel_physical_size_xyu
bf_dtype = slide_io.vips2bf_dtype(merged_slide.format)
out_xyczt = slide_io.get_shape_xyzct((merged_slide.width, merged_slide.height), merged_slide.bands)

ome_xml_obj = slide_io.create_ome_xml(out_xyczt, bf_dtype, is_rgb=False,
                                        pixel_physical_size_xyu=px_phys_size,
                                        channel_names=marker_names,
                                        colormap=None)
ome_xml = ome_xml_obj.to_xml()
# Save merged slide to `dst_f` with ome-xml
slide_io.save_ome_tiff(merged_slide, dst_f=dst_f, ome_xml=ome_xml)```

[cdgatenbee]

Glad to hear it worked out, @kriegy98! Thanks for providing the additional missing pieces that would go at the start of the script. In the next version of valis I'll be sure to update slide_io to address the warnings from scipy.