Dear Rohan,

I'm routinely doing tile imaging on slice using Leica SP5 or Zeiss 710 confocal microscope.

Did you checked that the bias in horizontality is not due to polish-nail/coverslip thickness? Because sometimes when you use a coverslip that do not spread all over the slide, you can have this kind of problems (see schema).You can correct it by adding a second coverslip to restore horizontality. If it's not the case, and if your yous slice is really not parallele, I advice you to acquire the slice as it is but to begin the acquisition with the more horizontal slde. Then you will be able to cut into your plane to erase the Z that are not present in all your stacks. Another possibility is to use Imaris software to realigned the stacks. Hope it can helps. Lydia.

Is it in backscattering mode? So to understand better, your images are slightly defocused on the one side? Or is it a different problem?

Are you implying a not parallel cut or squeezed/streched tissue? I would by all means not try to change this proportions but scan it as it is and using the data without re-shaping the cube. Try to cut differently with different knifes, vibration aid, ... To solve the cutting rather

With large pieces of tissue that we have imaged using a motorised XYZ stage to collect the whole section as a series of z-stacks, we can clearly see that some areas of the section are higher.

If I understand you correctly, you have series of Z-stacks at different X-Y positions. When you check vertical cross-section you see that at different stacks your tissue is at different level, so everything points to not-so-stable and repeatable stage.

I do not know any programs that allow to do correct this automatically, I'd suggest to write a page (or two) of code for MatLab (or whatever is comfortable for you). One of possible algorithms is:

1) extract a vertical cross-section from two neighboring stacks along the same line

2) find overlapped areas from different stacks and determine how much they are shifted with respect to each other and modify Z data of the z-stack

3) repeat this until all stacks are properly leveled with respect to each other

4) interpolate shifted z-stacks in order to have sections at the same z level everywhere

Hope this will help you somehow

Dear Rohan,

I'm routinely doing tile imaging on slice using Leica SP5 or Zeiss 710 confocal microscope.

Did you checked that the bias in horizontality is not due to polish-nail/coverslip thickness? Because sometimes when you use a coverslip that do not spread all over the slide, you can have this kind of problems (see schema).You can correct it by adding a second coverslip to restore horizontality. If it's not the case, and if your yous slice is really not parallele, I advice you to acquire the slice as it is but to begin the acquisition with the more horizontal slde. Then you will be able to cut into your plane to erase the Z that are not present in all your stacks. Another possibility is to use Imaris software to realigned the stacks. Hope it can helps. Lydia.

There is a free software called "ImageJ" that allow you to load, eventually re-size, and stack a series of images. Try with this. I leave you the link: http://rsbweb.nih.gov/ij/index.html

If it is an x-y missalignment job I agree with Andriy, and additionally you can use photoshop to sequentially align features manually. If not then I agree with Lydia; additionally however since in the second case it has to be a different focus problem I can propose an option that is a bit heuristic and might or might not work: if the two regions of different focus are distinct you could cut and translate the higher lying regions fluorescence down to the beginning of the lower lying tissue fluorescence so ta practically make them level and equi-focus if you can catch my drift. Of course this would be to salvage your existing stacked images, and it is a trick. Valid I would say but still a manipulation. I would start as the colleagues above suggested by trying to obtain flat and even samples.

I'd suggest looking at FiJi. It has a number of plugins to realign

stacks under Plugins->Registration. For example, SIFT, which aligns

features between planes. But depending upon the specific data, there

are quite a few options here, so it's probably worth taking the time to

look at what each plugin does, and see which is the most appropriate.

I've seen one used to reassemble tens of thousands of overlapping TEM

images into a single volume, so you're not restricted to stacks

either--some are restricted to simple stacks; others will reassemble

arbitrary sections.

Hi Rohan,

if I understand ur question correctly, u need to flatten ur images to stitch them into a collage in 2-D (x-y) and not 3-D (xyz). Then ur solution might be as easy as making maximum projections (a.k.a. Extended focus) of all ur images which will eliminate the focus problem and unevenness of ur different z stacks, granted the bulk of ur images structures are in ur images.... Hope this helps.

Huygens probably has a way to correct the stacks.

Thank you all so much for contributing . Andriy, is correct. Essentially, we have many (50-100) z stacks in separate x-y positions. Each of these stacks are combined to form a single image. Combining the separate stacks into a single image is not a problem. We can also obtain in a fairly straightforward way a maximum or minimum intensity projection image. I actually don't want to do this though. What I want is to do is to the render the stack in 3D -with its actual volumetric proportions intact. When we do this we can see unevenness in the rendered image. This may be from sectioning, or maybe from the labeling process - my instinctive feeling though is that is not because of the position that the slide is in while imaging. I will definitely check this out though.

Hi Rohan,

Have you tried Xuv stitch? stitch-http://www.xuvtools.org/

I routinely use this to stitch whole mounts of retina (which clearly haas bumps in Z)

The stitched image is then visualized in Imaris. Only issue is you will need loads of RAM.

If you have reasonable overlap you get pretty good images. Good Luck

I'm also involved in this project and would like to clarify a bit:

As Rohan said we are able to stitch tiles (a series of z stacks) together to form a 3D mosaic of a whole brain section - but then sectioning (probably) artifacts become apparent as we can see that the section wasn't lying flat on the slide, or was warped during specimen preparation, or etc. We are trying to minimise artifacts as we prepare the tissue but inevitably there is always some warping.

We are trying to implement a digital solution to correct for these artifacts, as it is virtually impossible to reconstruct an entire brain using sections that are all warped in varying ways. The solution we are trying to implement is outlined in the paper 'Three-dimensional reconstruction of serial mouse brain sections: solution for flattening high-resolution large-scale mosaics - Berlanga et al. (2011)'.

The authors of that paper used IMOD viewer, software commonly used in the electron microscopy community but works on light microscopy images. Basically our problem is that it is Unix based, and there are a number of technical issues stopping us using it (we use Windows machines, and so far attempts to use Unix-like environments have failed). The process involves placing many points on the image in 3D to define each warped boundary, and then a flattening algorithm is applied to transform the image in three dimensions to make each surface flat. This might be better explained visually, this website demonstrates the process - https://confluence.crbs.ucsd.edu/pages/viewpage.action?pageId=13042141#Howtoflattenvolumespecimens?-flat3DFlatteningawarpedvolumein3D

Is anyone aware of alternative software that can implement a flattening algorithm?