VALIS generate mask for the control tissue of the IHC, not for the wanted tissue

[Shentl]

Hi, thanks for your great work.
When i use VALIS to do registration between HE and IHC slides, i found that VALIS is very likly to generate masks for the control tissue in the IHC, and the truly need tissue will be ignored. How can i deal with it?

For example:

[Shentl]

i have set "create_masks=False"，but the result is same (registrar = registration.Valis(slide_src_dir, results_dst_dir, create_masks=False))

[Shentl]

The problem has been solved, in "preprocessing.entropy_mask()", the default "rel_min_size" is 0.2, but if the tissue in the IHC is too small, than it will be ignored

[Shentl]

Another problem，when the IHC slide have the control tissue，the width will be very huge. While in VALIS, it seems that the model will resize the longer side to a fixed value, which will the HE and IHC at the different magnification， which will makes the regastration fails

[Fuar25]

Have you solved the problem when the IHC slide have the control tissue? I'm having the same problem.

[cdgatenbee]

Hi @Fuar25 and @Shentl,
I have fixed the issue where settingcreate_masks=False would still apply the mask. I think it should resolve this issue, since it will use the entire image and not crop to the control tissue. That fix will be in the next update, which is just about ready, and I'll post here when it's been updated. In the meantime, maybe try using this image processor, which is the same as the default, except that the mask it creates covers the entire image, having the same effect as setting create_masks=False:

from valis import preprocessing
class NoMaskProceser(preprocessing.OD):
    """
    Use default ImageProcesser, but change `create_mask` to return
    a mask that covers the whole image. Should avoid cropping too
    much of the image or the wrong part of the image
    """

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def create_mask(self, *args, **kwargs):
        img_mask = np.full(self.image.shape[0:2], 255, dtype=np.uint8)
        return img_mask


# Create a Valis object and use it to register the slides in slide_src_dir, aligning *towards* the reference slide.
registrar = registration.Valis(slide_src_dir, results_dst_dir, reference_img_f=reference_slide, img_list=img_list)
rigid_registrar, non_rigid_registrar, error_df = registrar.register(brightfield_processing_cls=NoMaskProceser)

Please try that out and let me know how it goes.

Best,
-Chandler