Problem with Colocalization Threshold plugin and importing Olympus fluoview images

[Vera, Pedro L.]

Hello:
I am having a problem with determining what is the proper sequence of events when importing Olympus fluoview images and using the Colocalization Threshold plugin.

The images are 12-bit, but imageJ calls them 16-bit. I get widely different results depending on the sequence of events.

For example:

Sequence 1: Import (with LOCI)  and split channels
                    Convert to 8-bit
                    Adjust brightness-contrast (so I can see and pick a ROI)
                    Run Colocalization plugin

Sequence 2: Import and split
                    Adjust brightness
                    Convert to 8-bit
                    Adjust again (CRUCIAL step to stretch the histogram)
                    Run colocalization plugin

Sequence 1 yields very high (often perfect) colocalization which I can "see" is wrong. Sequence 2 matches more what I can see in the images. I know that under sequence 1 the LUT histogram remains very narrow, which is not the case with sequence 2.

I found and installed a fluoview plugin but I can't seem to get it to run (giving all kinds of java errors). At any rate I would appreciate any guidance on whether my empirical solution to this problem (sequence 2) is acceptable.

Thanks.

Pedro

[Daniel James White]

Hi Pedro,



Begin forwarded message:

> Date:    Fri, 30 Apr 2010 14:07:59 -0400
> From:    "Vera, Pedro L." <[hidden email]>
> Subject: Problem with Colocalization Threshold plugin and importing Olympus fluoview images
>
> Hello:
> I am having a problem with determining what is the proper sequence of event=
> s when importing Olympus fluoview images and using the Colocalization Thres=
> hold plugin.
>
> The images are 12-bit, but imageJ calls them 16-bit. I get widely different=
> results depending on the sequence of events.

the detectors analogue to digital converter is 12 bit,
but computer files are always stored in power of 2 bits,
usually, 8, 16, 32, 64 bits.
So a 12 nit images is always actually as 16 bit image, where the last 4 bits are not used.
Thats why imageJ opens it and decalres that its 16 bit.

>
> For example:
>
> Sequence 1: Import (with LOCI)  and split channels
>                    Convert to 8-bit
>                    Adjust brightness-contrast (so I can see and pick a ROI=
> )

ba careful here..... if you do that with different images, then
since theey might have different intensity levels , ir one sample was brighter than another,
with same detector settings, then after auot adjust, they will "look" the same brightness.

actually the convert to 8 bit already does that,
so you should turn that off and do it manually,
and use the set button to always set the range to convert to 8 bit as the same, eg 0-4095.


I have made a version of the colco threshold plugin that now works properly for 12-16 bit images...
it will come out in the Fiji distro of imageJ soon.

>                    Run Colocalization plugin
>
> Sequence 2: Import and split
>                    Adjust brightness
>                    Convert to 8-bit
>                    Adjust again (CRUCIAL step to stretch the histogram)
>                    Run colocalization plugin

maybe you can see here, now that you will end up with different results,
depending which way around and how you do the histogram stretch and to 8 bot conversion.
You should do your best not to lose the relative intensity info of the 2 colour channels
since it might be interesting (the slope of the regression line in the scatterplot)

to help get things done right,
see these 2 tutorial i made on the Fiji wiki

http://pacific.mpi-cbg.de/wiki/index.php/Colocalization_Analysis
http://pacific.mpi-cbg.de/wiki/index.php/Detect_Information_Loss

you need to look at the intensity histograms very carefully....

best would be to not convert to 8 bit at all,
but do the analysis on the original 12 bit data.

do the analysis on a region of interest not the whole images (esp if there are areas of black-ish background - you dont want to measure the background!)

do the Costes significance test also on the same ROI.



>
> Sequence 1 yields very high (often perfect) colocalization which I can "see=
> " is wrong.

maybe everything is above threshbold.... watch out for wrong 0 offsets.
Use an ROI.

> Sequence 2 matches more what I can see in the images. I know th=
> at under sequence 1 the LUT histogram remains very narrow, which is not the=
> case with sequence 2.
>
> I found and installed a fluoview plugin but I can't seem to get it to run (=
> giving all kinds of java errors). At any rate I would appreciate any guidan=
> ce on whether my empirical solution to this problem (sequence 2) is accepta
> ble.
>

see the tutorials above,
and if you still struggle  
just ask me.

Dan


> Thanks.
>
> Pedro=

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Visualisation, Processing and Analysis
Light Microscopy and Image Processing Facilities
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
[phone] (Work phone at MPI-CBG)
[phone] (Fax MPI-CBG LMF)

http://www.bioimagexd.net  BioImageXD
http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
http://www.chalkie.org.uk                Dan's Homepages
https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
dan (at) chalkie.org.uk
( white (at) mpi-cbg.de )