Hyperplexing multi-channel immunofluorescence images

[Doc Watson]

My apologies if this has been asked before, but I wasn't able to find
previous threads that really answered this.  Here's what I'm trying to do:

I have tissue sections imaged on a Zeiss slide scanner that produces large,
tiled and stiched, pyramid composites with 20x magnification, and 5
fluorescent channels (.czi format).  Then I'm doing pretty standard
cyclic-IF where I strip the fluorphores from the tissue, stain with new
markers, and image again.  This leaves me with 8-10 5-channels images of the
same tissue that I then want to align and merge together to create one image
with ~40 channels.

So I need to align every channel from each staining round to the images from
the first round of staining, and probably perform small local deformations
in case the tissue is warped during the stainings.  Ideally, I'd like to use
the DAPI channel from each 5-channel image to perform alignment and
registration, so that any image deformation is also applied to the other
channels from that round of staining.

I'm looking at using TrakEM2, but does anyone have other suggestions for how
to approach this?

And does anyone know a way to keep the pyramid structure of the .czi file,
or will I have to convert the whole thing to a BigTIFF image?

Thanks!



--
Sent from: http://imagej.1557.x6.nabble.com/

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Kenneth Sloan-2]

Everything depends on the details, but...

I would start by deciding what flavor of transformation you
actually need.  I would avoid even thinking about local warping
until and unless you find it is absolutely necessary.

Assuming that the tiling done by the Zeiss slide scanner is
adequate, I would create a low resolution overview image for
each staining step and align these.

Sorry - I don't know how to communicate the resulting transformations
to the full .czi pyramids.

Without knowing what you really need to do, I might also consider
processing the 11-deep stacks tile-by-tile.

If you NEED the full image, then perhaps lower resolution is adequate, and
you don't need a full pyramid.  On the other hand, if you need the finest
available resolution, processing tile-by-tile may allow you to compensate
for more complicated warping of the tissue (each tile might be adequately
registered using a simple transformation.

Finally, you might do the high-resolution work on the individual
(smaller) stacks, and later register a higher-level, more abstract
representation of the results.

I don't personally use TrakEM, but I have colleagues who swear by it.
In a somewhat similar situation (serial section and block-face EM),
they produce .obj models of what they see at high resolution, and
then ask me to register and combine the .obj models.  
This works surprisingly well.

Above all - it's important to have a model of the CAUSES for
mis-alignment.  Is it imaging?  tiling? tissue warping?  Different
causes call for different fixes.

--
Kenneth Sloan
[hidden email]
Vision is the art of seeing what is invisible to others.

> On Feb 24, 2019, at 07:21, Doc Watson <[hidden email]> wrote:
>
> My apologies if this has been asked before, but I wasn't able to find
> previous threads that really answered this.  Here's what I'm trying to do:
>
> I have tissue sections imaged on a Zeiss slide scanner that produces large,
> tiled and stiched, pyramid composites with 20x magnification, and 5
> fluorescent channels (.czi format).  Then I'm doing pretty standard
> cyclic-IF where I strip the fluorphores from the tissue, stain with new
> markers, and image again.  This leaves me with 8-10 5-channels images of the
> same tissue that I then want to align and merge together to create one image
> with ~40 channels.
>
> So I need to align every channel from each staining round to the images from
> the first round of staining, and probably perform small local deformations
> in case the tissue is warped during the stainings.  Ideally, I'd like to use
> the DAPI channel from each 5-channel image to perform alignment and
> registration, so that any image deformation is also applied to the other
> channels from that round of staining.
>
> I'm looking at using TrakEM2, but does anyone have other suggestions for how
> to approach this?
>
> And does anyone know a way to keep the pyramid structure of the .czi file,
> or will I have to convert the whole thing to a BigTIFF image?
>
> Thanks!
>
>
>
> --
> Sent from: http://imagej.1557.x6.nabble.com/
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html