Hello Aldo and Lenny
I, too, would suggest one more try of Coloc 2. If you find any problems,
please report them. Though, he plugin isn't really finished, yet.
However the calculations should be robust, only the user interface is
not as pretty as the JACoP one (yet) :-).
On 20.06.2012 06:01, Leonardo Guizzetti wrote:
> Perhaps I can help and point you back toward the Coloc2 plugin in
> Fiji/ImageJ. I am doing colocalization of proteins using confocal
> microscopy and turned initialially to the Colocalization plugin within
> ImageJ (using the MacMaster University biophotonics distribution which is
> quite stale). Getting into it, the program did not truly handle ROI's, and
> as well reported numerically incorrect values for Pearson's. I ended up
> helping debug the Coloc2 plugin (which was also not properly handling
> ROI's)
Indeed, there where some ROI and mask problems in the beginning of Coloc
2. However, these are fixed by now and the mask and ROI handling should
be working. (At least all the unit tests are green :-).)
You could also have a look on the wiki page on fiji.sc [1] for some
general notes on Coloc 2. JACoP and the other plugins. This page needs
some updates, though.
> Long story short, the Fiji Coloc2 plugin does:
> - correctly handle ROI's,binary masks, and full images for all parameters
True and to add some more words about these: ROI's can be any type of
ROI ImageJ/Fiji offers. However, they always will be the same for every
slice. If you want to have a variation in Z (for a Z stack) as well, you
can use the masks that Lenny mentioned. You just need to create an image
with the same extent as your original one. Then everything with a zero
value (back) will *not* be taken into account, everything else will.
Duplicating your stack and using thresholding and make it binary, could
support you in marking the structures you are interested in.
> - Pearson and Mander's coefficients (M1, M2)
Yep.
> On Tue, Jun 19, 2012 at 11:56 AM, Aldo Vacaflores wrote:
>> Also, I tried using Coloc2 which has the option to select an ROI's
>> but I can’t figure out what value to use when it asks for the PSF.
> - PSF is only needed for another of the calculated parameters, not
> these two in particular.
Yes, but it might be useful to investigate it. The algorithm it is used
for is the Costes statistical significance test. It indicates whether
your Pearson coefficient is not better/more significant than one
obtained between your images if one of it was randomized/shuffled. That
in turn means that your Pearson coefficient is not significant. You can
find a link to the paper on the Fiji wiki. Now this shuffling is done by
randomizing blocks of the original image. The length of such a block's
edge is equal to the point spread function (PSF). So either you take the
PSF of your original image/microscope setting or you leave it with the
default (which is in most cases fine as well).
Also note that when choosing an irregular ROI or mask, our
implementation of Costes shuffles the bounding box around it. Of course,
image data is only used from within the ROI, but the remaining parts are
black and still shuffled around.
> Right now, the advantage that JaCoP plugin has is a prettier interface.
True. But one advantage of Coloc 2 is that it is tested (in terms Java
unit tests) . At least as far as I know is this not done with the JACoP
plugin. Unfortunately, we didn't find time yet to work on the user
interface.
> I will also advise using Pearson's over M1 and M2, as per the
> recommendation of Parmryd & Adler. ( 1. Adler, J., and Parmryd, I. (2010)
> Cytometry. Part A : the journal of the International Society for Analytical
> Cytology 77, 733-42 [online]
http://www.ncbi.nlm.nih.gov/pubmed/20653013. )
Thanks for posting, I didn't know this publication.
Also, hopefully in a few days an update of the Coloc 2 plugin will be
available which includes some bugfixes (and is based on Imglib2). So
have your Fiji up to date :-).
Cheers,
Tom
[1]
http://fiji.sc/wiki/index.php/Coloc_2--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html
Locked
Jun 19, 2012; 3:56pm
