Chromatic shift correction

[Rafael Buono]

Dear list,

I am working on a dual camera spinning disk system and would like to use ImageJ to correct for chromatic shift between the two channels (the two cameras).
I have data from fluorescent beads and would like to use this data to calculate the shifts (and potential warping) between the channels and apply the necessary transformations to my data (dual channel, z-stack, time course datasets). Is there a plugin that can do this?

I have read the entry on imagej.net ( http://imagej.net/Chromatic_shift_origins_measurement_and_correction ), but I have not found an automated way to get ImageJ to use the bead data to apply corrections to my actual data. How are you handling this type of problem (having to correct for chromatic shifts between channels)?

Best,

Rafael Buono

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

[Saalfeld, Stephan]

Hi Rafael,

we developed such a method recently and I generated a slightly
premature screen capture how to do it.  Instead of polishing it
further, we decided to upload the screen cast to Youtube

https://www.youtube.com/watch?v=lPt-WQuniUs

and link the relevant scripts and examples on GitHub

https://github.com/saalfeldlab/confocal-lens/tree/master/scripts

You will need a dense textured sample that looks approximately the same
over the entire spectrum and image it as a 4x4 mosaic of ~60%
overlapping tiles.  We used beads.

Best,
Stephan


On Mon, 2016-11-28 at 15:08 -0500, Rafael Buono wrote:

> Dear list,
>
> I am working on a dual camera spinning disk system and would like to
> use ImageJ to correct for chromatic shift between the two channels
> (the two cameras).
> I have data from fluorescent beads and would like to use this data to
> calculate the shifts (and potential warping) between the channels and
> apply the necessary transformations to my data (dual channel, z-
> stack, time course datasets). Is there a plugin that can do this?
>
> I have read the entry on imagej.net ( http://imagej.net/Chromatic_shi
> ft_origins_measurement_and_correction ), but I have not found an
> automated way to get ImageJ to use the bead data to apply corrections
> to my actual data. How are you handling this type of problem (having
> to correct for chromatic shifts between channels)?
>
> Best,
>
> Rafael Buono
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

[Rafael Buono]

Hi Stephan,

Thank you for the prompt answer, the helpful video, and the cool scripts. I will give it a shot with the next wave of data.
A couple of questions:
-Does it need to be done on 4x4 tiles or is it only if I am going to be doing tilling with my data?
-How do I apply the transforms obtained from the bead data to my actual data?

Best,

Rafael

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

[Saalfeld, Stephan]

Hi Rafael,

The 4x4 tiles are required to calculate the distortion, so the answer
is yes, you will need to do this, but then you have a calibration for
this particular setup of your microscope that can be re-used for all
subsequent acquisitions, tiles or not.

Documentation is sparse to the level that it does not exist, so let's
start here.  Following an e-mail that I sent out to the Fly Light
project scientists who are testing this:

After you followed the instructions from the video, check this place
again

https://github.com/saalfeldlab/confocal-lens/tree/master/scripts

and open the script export-transforms.bsh in Fiji's script editor and
modify

https://github.com/saalfeldlab/confocal-lens/blob/master/scripts/export
-transforms.bsh#L13-L21

according to your setup.  Hit Run.  It should print the estimated
transformations into the log window and you can save this as a JSON
file (text file).

Now drop the file apply-lens.bsh into the plugin directory of your Fiji
installation.  After a restart of Fiji, you should find a menu-entry
for this script in the plugins menu.  Click it.

If your Fiji is not too old, then it should open a dialog that asks you
for two input image paths (expected is

1. LSM file with two channels (488 and 594nm)
2. LSM file with three channels (488, 561, and 647nm)

), the respective JSON file for the scope that was used, and an output
directory to save the resulting multi-channel file.  The output will
have all five or however many channels in a single file.

If your acquisition procedure is different, change the script.

The script currently assumes that the lens-models in the JSON file and
the channels from input images are in the same order (in this case 488,
594, 488, 561, 647nm).  You can do subsets or other combinations but
you would have to edit the JSON file accordingly.  The JSON files are
simple to read, check

https://github.com/saalfeldlab/confocal-lens/blob/master/scripts/scope1.json

and they contain name tags for each model.  That should make editing
them easier.  The script ignores the name tags and goes by the order
only.

Attached is an example showing the top left 200px corner of all five
channels of one stack of the scope 1 calibration sample before and
after correction.

The script should be macro-recordable.


 On Tue, 2016-11-29 at 15:55 -0500, Rafael Buono wrote:

> Hi Stephan,
>
> Thank you for the prompt answer, the helpful video, and the cool
> scripts. I will give it a shot with the next wave of data.
> A couple of questions:
> -Does it need to be done on 4x4 tiles or is it only if I am going to
> be doing tilling with my data?
> -How do I apply the transforms obtained from the bead data to my
> actual data?
>
> Best,
>
> Rafael
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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