Hi,
I've been using CMTK to register functional 2-photon calclium imaging stacks of larval zebrafish to a template. While I am in general happy with the registration result, I noticed there are some weird artifcats in the images, where you see a moving stripe or wave of brightness that wasn't there in the original. It's a bit hard to explain so I'll upload a short movie where I go through the stack and put the mouse where this moving stripe appears. I've seen this or similar patterns in three different stacks I've registered and I wanted to find a way to register them without creating this artifact. I tried applying -match-histograms or --histogram-equalization-flt but it didn't change anything. Has this happened to you? Do you have any suggestions on what parameters to tweak to change it?
Thank you,
Manuel Bettencourt
Hi Manuel -
This is just a wild guess, but it looks to me like there might be, in your original data, one slice that is a little brighter than the remaining slices. When you reslice your data, say using a slight rotation, then the brighter slice appears to "travel" through the resulting stack as you scan through its slices.
You might be able to confirm this by scanning through your original data perpendicular to the brighter slice, but it might be too subtle to spot. If you have some way of doing a projection of your 3D stack onto a 2D plane (e.g., by maximum intensity project, or by averaging) then that might make the effect more obvious and help confirm this theory. I vaguely remembered that CMTK might have something to do that, but cannot find it right now, so it's possible that has been removed over the years.
Be that as it may, CMTK does have a "destripe" command line tool that might be able to correct this effect. Basically, its purpose is to identify slices in your data that deviate from the brightness of their neighbours, and adjust them accordingly. I haven't used this in well over a decade, obviously, so I don't have any specific suggestions for how to use the tool. But it might be fairly straightforward, so maybe give that a shot.
Best, Torsten
Hi,
I think you are right. Indeed, as one can see from the coronal average intensity projection, there are planes which are brighter than others (see annexed png). However, I couldn't solve this with destripe. It seems to make the bright slices less different from the others but you can still see them in the projection image and the artifact still stays after the registration. I tried to play around with kernal-fwhm and I've been using the code below. I'll keep trying to tune this, but thank you very much for your reply.