Hi,

I have been trying to register a number of brains with distinct Gal4 expression patterns.

However, the resulting reformatted brain gives strange results (see attachment).

Could someone please explain why this happened and how I can avoid this in the future?

Thank you!

Hi Kiliana -

Indeed, that looks a little odd. But in order to make a guess what could be the cause of this, it would be helpful if you could post your warp command line.

Also, could you post a picture (as a PNG or JPEG, not the whole 3D stack) of each of the two images involved? Perhaps a central z slice?

Best,
  Torsten

Hi Kiliana,

To add to what Torsten has said, you should post something so that we can see the image that was used for registration. I am guessing that the template (reference) image was the IS2 template brain:

http://flybrain.mrc-lmb.cam.ac.uk/si/bri...

But we would need to see the image that you used for registration. A Z projection or single slice would be a good start, but it might be a good idea to put the whole stack online with google drive/dropbox/etc. One common problem is if your brain is rotated a large amount or flipped with respect to the template. Note that the channel 02 output image that you sent to us is the signal channel and not the nc82 reference channel that was used for registration.

In terms of the warp command line that Torsten mentions, you will need to send either the manual command line or the script created by the Fiji CMTK gui, which will be called something like munger_2015-08-20_12.05.53.command.

Best wishes,

Greg.

Hi,

All the data I used for reformatting are available on google drive (https://drive.google.com/folderview?id=0...)
In addition I included a JPEG of the Z-stacks of the original images.

The registration was performed with channel 1 as an nc82 staining and channel 2 as a gfp staining. This resulted in two reformatted images, but these are basically twice the same distorted image stack.

Best wishes!
Kiliana

Hi Kiliana,

I took a look at your data.

1. the slice order of your stacks is back to front with respect to the template brain – the registration software needs the floating (sample) image to be roughly aligned with the template before starting.
2. This is probably because you have mounted your brains anterior face down onto the slide
3. When you did this, you probably didn't use a spacer between the coverslip and the slide – the brains look very squashed. A broken piece of #1 glass coverslip is about the same width as the brain (150-170 µm)

Although it is surprising what Torsten's registration software can achieve I think you will probably never have much luck with these particular specimens. You may want to take a look at

* http://cshprotocols.cshlp.org/content/2013/4/pdb.prot071738
* http://cshprotocols.cshlp.org/content/2013/4/pdb.prot071720

For some more details on specimen prep required for registration.

Best wishes,

Greg.

Hi -

To add to what Greg said:

You COULD try to make this registration work by editing the header of your images and increasing the slice thickness to approximately match the thickness of the template brain. The result would be fairly bad, though, since you'd still have only 14 slices to cover the entire brain.

You COULD also get around the upside-down problem by initializing the first registration stage (linear affine) with a 180 degree rotation. If you did that, you could also, instead of editing the image header, provide an initial scaling factor in z direction to "unsquish" the image. (The advantage of editing the image header would be that it should work regardless of your processing, i.e., regardless of whether you use CMTK tools directly or wrapped inside Greg's munger script).

Finally, your two image channels are rather different, and in all likelihood one of them (my guess, the first) will be giving a lot better registrations than the second, because it is more similar to the staining that went into the reference brain. This difference would not necessarily have much effect in the linear registration, but certainly in the nonrigid warp.

Best,
  Torsten

Hi,

Thank you very much for your input. I will follow up on your suggestions and generate some new samples.

Would it be possible to obtain a pdf version of the protocol via e-mail? Unfortunately I cannot access the articles mentioned, and these protocols would be really helpful for my sample preparation.

With regards,
Kiliana

You can access these at:

https://www.researchgate.net/publication...
https://www.researchgate.net/publication...

and for good measure

https://www.researchgate.net/publication...

but you may need to log in first.

Best,

Greg.