I'm trying to use the BrainAligner to register the brain images I got. But the Janelia website hosting the manual is down.
 
I would really appreciate if you can give me an instruction of how to use it, such as which commanders I can use and what kind of images I should have.

Hi Kevin,

Thanks for interesting in our alignment tool. You may refer to Dr. Hanchuan Peng's paper https://www.nature.com/nmeth/journal/v8/... to download the software or compile from the source by your own from https://github.com/Vaa3D/v3d_external/tr... Basically it is a landmark-based alignment method, you need a template image and a landmark set of it (you can create it using vaa3d manually or detect it automatically with your own methods), then try running it with command line: 

jba -t "template_image.tif" -s "subject_image.tif"  -L "template.marker" -w 10 -o "output_image.tif" 

The JBA aligner is developed for aligning general 2D/3D image alignment. Sometime we might need pre/post process images to let the alignment work as we expected. It is highly depending on your studying cases. Hope this will help. Feel free to contact us if you still have any questions.

Best,
Yang 

Originally posted by Kevin Leo:

Hi Yang,

Thank you so much for your information!

I finally got the vaa3d brainaligner working:


I want to align my fly brain images to a template fly brain (the JFRC2 template: zenodo dot org/record/10567#.WfnFlUyZMo8) and I found that it does not work as I expect. I sometimes got images with no alignment, or empty images, or flying brains...

Can you please tell me which steps are very critical?

I really appreciate your help!

Hi Kevin,

Looks like you have had a great progress.

The JBA can handle both local (non-linear) and global (linear) alignment problems but from my experience I will rely on other open source toolkit (ANTs/CMTK) for the global alignment.

Usually, we will create a pipeline for a bunch of similar images need to be aligned.

1) preprocessing step: I will process the subject image to make its orientation as the same to the template's due to most alignment software can not align images with bigger rotations especially having flips along the axis in some cases. Using some prior knowledge of the resolution info, we can make the subject image's scale the same to the template's.

2) global alignment to get affine matrix (you can merge previous step into this affine matrix as well)

3) local alignment to obtain the corresponding landmarks found in your subject image.

4) warp your subject image based on the transformations obtained from above 3 steps.

At Janelia, we develop a bunch of pipelines for different cases. If you need further help, please share your example image with us. I will find time to look into it.

Best,
Yang

Cool! Thank you so much, Yang! This is very informative. I'll try it by myself first and will let you know how it goes.

My purpose is to map the expression of several different transcription factors into a single brain. But the problem is that each antibody against different transcription factors has different background. Do you have any suggestions about how to reduce the background?

Another issue of my images is that the signal is getting weak when the location is getting deeper. There is a signal decay along the Z axes. Do you know there are any methods I can use to adjust it?

Many thanks!

Hi Kevin,

For the background, without looking at your images, I can only give some general ideas. You can use adaptive thresholding (rolling ball) to separate foreground and background signals. If you are interested in some fancy algorithms, you can train a model to classify your image's pixels into foreground and background.

The "bleaching" problem, you can try some image enhancement methods to normalize the image contrast. However, most researchers try to enhance the label effect from biological aspect. I am not sure why you still have this issue. I would suggest you try improving your imaging instead of post processing.

Hope this will help.

Best,
Yang