Hello,
I am trying to register a high resolution H & E stained image(RGB) on a segmentation image.
The RGB image size is: (620, 734, 3)
and label image is: (96, 73)
The problems are:
Is it possible to register 3 channel RGB image on the label map? In this case shall grayscale image be used or taking a single channel from the RGB?
what kind of resampling, spacing, and dimension of both fixed(label) and moving(hne) should be considered?
I have uploaded the demo data here: hne_label.mat (1.4 MB)
That is a strange problem. Segmentations are usually done using some “feature” image as a reference. And you register feature images, and apply the resulting transform on the segmentation.
Using information from multiple channels during registration is hard. I suggest converting your RGB image into grayscale.
Before writing any code, I suggest using Slicer, and its BRAINS and Elastix registration modules to experiment.
The segmentation map/image comes from a different modality. And I am trying to fuse both of them into the same space. That is why I am trying to register hne on label. Also not having an atlas makes more problematic.