Scaling issue

[LolaaPe]

Hi Johannes, When running S2V with the attacched parameters, the Output Stack doesn't fit the atlas I registered my histological images to (image attached). The output stack seemes to be much smaller than the atlas. Thanks for your help!

[jo-mueller]

Hi @LolaaPe ,

thanks for reporting! I encountered this error earlier and thought I had it fixed for most cases, but apparently that's not the case. Two questoins:

• When you run S2V on histological data, you should be presented with a dialogue - something like Image is much larger than target - proceed with downsampled image?. This you have to confirm, otherwise it will not work.
• Secondly, does this occur for all slices or just for a few ones?

Things you could try:

• Try increasing the histo outline smoothing degree and the Volume outline smoothing degree (in the S2V publication I used values up to 10 or 12).
• If you look into the elastix_parameters.txt, there are these three lines:

(NumberOfResolutions 5)
(FixedImagePyramidSchedule 64 64 16 16 8 8 2 2 1 1)
(MovingImagePyramidSchedule 16 16 16 16 8 8 2 2 1 1)

You could try increasing the NumberOfResolutions parameter to 6, in which case you would have to change the two other lines into something like this:

(FixedImagePyramidSchedule 82 82 64 64 16 16 8 8 2 2 1 1)
(MovingImagePyramidSchedule 82 82 16 16 16 16 8 8 2 2 1 1)

Basically, this blurs both the fixed and the target image (the atlas image and the histological image to be registered) A LOT so that they overlap better - the registration is then able to do the rest. However, if they do not overlap well in the beginning, elastix doesn't know how to change the transformation in order to make the registration better - hope this makes sense.

[LolaaPe]

Hi @jo-mueller,

thanks for all your suggestions! Unfortunately nothing worked so far and I have discovered that the scaling issue is caused by the type of .tif files I am using to perform the registration. Due to the large size of my scanned brain tissue slides (20x magnification) Fiji is unable to open middle brain tissue slices as a whole image. That's why I extracted single-channel DAPI.tif images from Zen blue 2.3 via Image Export -> TIFF. The export function is also restricted which is why I chose to export the single-channel DAPI.tif without compression but applying 8-bit conversion and 'shift pixels'. I ran S2V with the 8-bit images but also changing the 8-bit DAPI.tif back to 16-bit in Fiji. Scaling issue sadly remains in both cases. When looking at the meta data of the DAPI.tif exported from ZEN I saw this: The metrics of the image appear to be in inches. I am not sure whether the meta data of the .tif files are responsible for the scaling issue? Nevertheless, have you performed S2V registrations on brain tissue slides that have been scanned with a 20x objective? Do you have a suggestions how I can extract single-channel DAPI.tif from my big images? Thanks!

[LolaaPe]

Solution: https://forum.image.sc/t/splitting-large-multi-channel-czi-to-single-channel-tif/85291/2

[jo-mueller]

Thanks for linking the solution here @LolaaPe ! (And sorry for not getting back to your last comment, pretty swamped right now 🙈 )