Hi!

I have been running registrations successfully on brains stained with nc82 but am failing to register brains labeled with N-cad. When I try register brains stained with N-cad, I get one of the two following errors:
1. There is a single reformatted image but it has nothing but small circular blob
2. Registrations failed.

Please advise. 

Ahana.

Dear Ahana,

Sorry to hear you're having trouble, but I'm afraid we'll need more information to help diagnose.

In case (1) are you saying that all the expected images are generated but they contain incorrect results or are some outputs missing.

In case (2) what error messages do you see and what output files are present. If you zip up a folder containing a couple of images/registrations and share it with me off list, I can try to diagnose.

All the best,

Greg.

Hi! 

Thanks for your prompt response. 

(1) I am trying to run registration in 3 separate channels as 3 separate images. I just get one image with a small portion of the whole brain aligned -- I am not really sure what it is but I can send you the resultant image. 

(2) I see "Run 2 affine registrations of which 2 failed". I actually have 3 input files. I can send it to you as a zip. My email is amallick@umd.edu. You can drop me an email and I can send you the files in that address. 

Ahana. 

Originally posted by Greg Jefferis:

Dear Ahana,

As we have discussed off list your problems stem from differences in image quality and also a spatial miscalibration issue.

I just wanted to note a point that may be helpful for others. The registration software itself generates a "pyramid" of downsampled image stacks, first aligning coarsely with the most heavily downsampled image data and then more and more precisely. If you have imaged at high resolution (e.g. less that 0.5µm in any axis for a fly brain), you can tell it not to use the original image data using the --omit-original-data argument.

For example you can change your call from:

"/Applications/Fiji.app/bin/cmtk/munger" -b "/Applications/Fiji.app/bin/cmtk" -a -w -r 010203 -X 26 -C 8 -G 80 -R 4 -A '--accuracy 0.4' -W '--accuracy 0.4' -T 2 -s "/Users/ahana/Desktop/registrationfolder/refbrain/refbrain.nrrd/images" .

To:

"/Applications/Fiji.app/bin/cmtk/munger" -b "/Applications/Fiji.app/bin/cmtk" -a -w -r 010203 -X 26 -C 8 -G 80 -R 4 -A '--accuracy 0.4 --omit-original-data' -W '--accuracy 0.4 --omit-original-data' -T 2 -s "/Users/ahana/Desktop/registrationfolder/refbrain/refbrain.nrrd/images" .

Note that:

-A '--accuracy 0.4 --omit-original-data'

Passes additional arguments to the initial affine registration. And

-W '--accuracy 0.4 --omit-original-data'

Does the same for the second nonrigid (warping) step. There is rarely any reason to use full resolution image data for the affine step but there could be arguments for leaving it for the warp.

All the best,

Greg.