Cell segmentation mask doesn't fit image

[pikapika505]

As the automatic cell segmentation did not work in SAW count, we manually labeled cells in Napari using the initial SN.tif (DAPI) and SN_IF.tif (FB28_IF) images. In Napari, all three images overlap and all files have the same sizes.

In the StereoMap4 Image Processing step, I loaded the SN.tif DAPI and SN_IF.tif FB28_IF files, and then opened the manually labeled SN_mask.tif file. However, the mask appears smaller and shifted relative to the images.

Then I registered images according to 254 issue and did:

path/saw/saw-8.2.1/anaconda/bin/python path/saw/saw-8.2.1/lib/register/register/main.pyc
 -i SN_DAPI.tif,SN_IF_mask16.tif 
-o . -w False --sn D05508E2 
-v path/outs/feature_expression/D05508E2.raw.gef 

This generated SN_DAPI_regist.tif. I then used SN_DAPI_regist.tif in StereoMap, but the mask still appears smaller and shifted.

What should I do to properly use my cell segmentation mask?

Another thing I noticed is that although each cell was manually drawn and labeled in Napari, and the .tif file was converted to a binary .tif because StereoMap only accepts binary mask files, the cells appear merged in StereoMap.

Why does this happen?

[Jinghao-Chen]

Hi, all images ultimately need to be aligned to the expression matrix. Therefore, if cell segmentation is performed using the original images followed by registration, issues such as flipping and scaling will occur, which may even lead to the deformation of some cell contours.

Thus, it is recommended to conduct cell segmentation with SN_DAPI_regist.tif first, and then import the generated mask.tif into StereoMap Step 4, rather than using the original images.

[pikapika505]

thank you for your prompt response @Jinghao-Chen

Our collaborators did manual labeling of thousands of cells already on the original DAPI and IF images. Is there any way to avoid doing that again and use the existing mask file?

[Jinghao-Chen]

Sorry, there is no convenient solution available at the moment. You may try using Fiji/ImageJ to register mask.tif to SN_DAPI_regist.tif, which will give you the registered mask.tif.

I remember your data has undergone manual registration and the SAW realign pipeline. When using ImageProcessing in StereoMap, please import the .stereo file output by the SAW realign pipeline, and load the registered mask.tif at Step 4. And SN_DAPI_regist.tif should also come from the SAW realign pipeline.