Aligning DAPI/mask images to regist.tif

[pool-es]

I am working in StereoMap and would like to improve the segmentation results using a cell segmentation mask from an external segmentation tool.

The manual strongly advises to use the regist.tif file as input for third party segmentation. However, using the regist.tif file results in inaccurate cell segmentation with third party algorithms, since the DAPI staining in the image is too bright to distinguish individual cells.

Ideally, I'd like to use the microscopy DAPI image file, with a different brightness, for third party segmentation. However, this DAPI image and the extracted third party cell segmentation masks do not match the size/orientation of the regist.tif. Therefore, I cannot overlay this third party segmentation mask on tissuemask in StereoMap appropriately.

I could register the original DAPI image in StereoMap and perform downstream segmentation, but feel that the segmentation results are better when loading the microscopy DAPI image file for external segmentation directly.

Is there any way to transform the original DAPI image (and third party masks derived from it without tracklines) to match the size/orientation of the regist.tif file? I've tried doing it manually in ImageJ or StereoMap, but it's not completely accurate and quite sensitive to mistakes. The x- and y- offsets seem to be the most difficult part. If I'm correct, the information on the regist.tif file is stored in the .ipr file, could I use that?

Many thanks for your help.

[Jinghao-Chen]

Hello, I'd like to confirm: is regist.tif too bright, or is the original DAPI image input for SAW analysis too bright?

[pool-es]

Thanks for the quick reply.
The DAPI image that was loaded into the saw_count run was too bright, this was done to make the tracklines visible. The original DAPI image is fine: the scaling of the DAPI image can be adjusted to reduce brightness.
I can load this DAPI image with reduced brightness into the image processing pipeline in StereoMap, align it with the matrix and then use this DAPI image for external segmentation. However, these segmentation results are inferior compared to the segmentation results when using the microscopy DAPI.czi file directly.

Therefore I was wondering if there are any direct ways to resize the external segmentation mask files to the regist.tif file? (or find the offset, rotation etc of the regist.tiff tile). Thanks again for your help.

[Jinghao-Chen]

Hello, you could try the following method:

1. Name the DAPI image that was loaded into the saw_count run as SN_DAPI.tif. SN is the chip number.
2. Then, name the TIFF images for which you need to adjust the brightness as SN_IF_Test.tif. You could name the 'Test' string part arbitrarily, but it must follow the format SN_IF_{String}.tif. If you have multiple TIFF images requiring brightness adjustment, you could also name and distinguish them using the same format.
3. Run the command:

/path/to/saw-8.2.0/anaconda/bin/python /path/to/saw-8.2.0/lib/register/register/main.pyc \
   -i SN_DAPI.tif,SN_IF_Test.tif \
   -o ${output_dir} -w False \
   --sn {SN} \
   -v /path/to/saw_output/outs/feature_expression/{SN}.raw.gef

The -o parameter is for entering your desired output directory;
the -v parameter is for entering {SN}.raw.gef in the outs/feature_expression directory under the output directory of the saw count;
the -i parameter is for entering the image files with the modified naming format.

After the run is completed, you will obtain the regist.tif image of the input images in output_dir.

[pool-es]

Thank you for this suggestion! It works perfectly, thank you.