Posted by
Vincent Schoonderwoert on
May 24, 2013; 2:05pm
URL: http://confocal-microscopy-list.275.s1.nabble.com/Colocalization-on-Z-stack-images-tp7580319p7580359.html
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy*****
**vendor response**
Dear Marjan,
Of course we are tempted to reply to your questions by focusing
specifically on our Colocalization Analyzer. Instead, we would like to
state that probably all well-known imaging software developers used the
originally published equations for calculating the specific
colocalization coefficients (like for example Manders, Pearson,
Spearman), and/or can tell you how they are used. So, the outcome of
these equations are very likely similar by the different softwares. But
there is more to it than just these different equations. That is why we
think your question should also be addressed by discussing the
additional imaging factors that influence colocalization analysis, and
how you can deal with these (see next paragraph).
First, however we need to mention that colocalization is not always the
best method to prove an obvious localization question. In your specific
case, you want to relate the location of objects in one channel towards
an object in the other channel. Coefficients like intersection or
overlap seem to be the most well-suited for showing this type of
"overlap", since correlation coefficients like e.g. Pearson and Spearman
focus more on the correlation of the signal in- and de-crease between
channels. Also, all these coefficients will take the fluctuating signal
values into account if the nucleus is unevenly stained, in which case
there will be no linear relation between the volume of the RNA objects
and that of the nucleus. Consequently, the coefficients will not give
you an accurate indication of how much RNA is within your nucleus and
how this changes under different conditions. Instead, you may consider
defining the nucleus as one volume using a ROI option within a 3D object
analysis tool, and subsequently measure change in volume of the RNA
objects with respect to the total volume of this ROI. I'll be happy to
discuss this approach off-line with you.
Back to your question on the quality of colocalization results.
These highly depend on the quality of image acquisition. Images should
be acquired as Z stacks - as you pointed out nicely, and they should be
properly sampled according to Nyquist. Undersampled images lead to
incorrect colocalization values, as the image is lacking detail.
Clipping/saturation of signal should also be prevented. Logically, these
two aspects cannot be solved with imaging processing software but need
to be dealt with during image acquisition.
Background signal also affects the colocalization coefficients, and can
- in most cases - be dealt with in softwares. Typical background
estimators, which details have also been published are Gaussian minimum
and Costes, or the optimized Costes method that does not assume the
background threshold combination is on the 2D histogram regression line.
See also:
http://www.svi.nl/ColocalizationCoefficientsBlurring by the optics, the structure of the blur (point spread
function), the introduction of noise, chromatic aberration (shifts
between channels), spherical aberration, and crosstalk. All can severely
affect the coefficient values.
Ideally, they should be minimized optimizing the imaging setup. However
software tools can efficiently deal with these factors.
On these web-pages you can simulations and more info:
http://www.svi.nl/ColocalizationBasicshttp://www.svi.nl/BlurAndNoiseAffectColocalizationPost acquisition correction methods:
Blurring, noise, and spherical aberration can be corrected with
deconvolution, preferably an iterative maximum likelihood algorithm
which is very suitable for noisy (confocal) images.
Asymmetry in the PSF structure can be corrected with deconvolution using
an experimental PSF distilled from beads.
Noise can also be corrected with RBNCC (see this paper by Drs. Jeremy
Adler and Ingela Parmryd
http://uu.diva-portal.org/smash/record.jsf?pid=diva2:563664)
Chromatic aberration can be measured, and corrected by shifting channels.
Crosstalk/bleed-trough by un-mixing tools
We offer tools for correcting these imaging artifacts, and you are
welcome to contact us off-line for additional questions.
Kind regards,
Vincent
Vincent Schoonderwoert, PhD
Senior Imaging Specialist/Account Manager
Scientific Volume Imaging bv
Tel: + 31 35 646 8216
www.svi.nl
On 05/21/2013 03:13 PM, Marjan Gharagozloo wrote:
> *****
> To join, leave or search the confocal microscopy listserv, go to:
>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy> *****
>
> Dear Philippe, Julio, and all confocal members,
>
> Thank you so much for your kind reply and valuable information. I'd
> like to know is Huygens professional is good for 3D colocalization
> analysis. It's the first time that I'm using it, it seems good but I
> don't know if the data is reliable enough for publication.
>
> Many thanks
> Marjan
>
> On 5/19/13, phil laissue <
[hidden email]> wrote:
>> Dear Marjan,
>>
>> apologies for the delay, I have been away on annual leave.
>> It is hard to tell from your description what might work best, since I
>> don't know what the structures are. In any case, you need something
>> that gives you good values in 3D. It is tempting to flatten the
>> z-stack, then do a Pearson's or something similar, but that is bad
>> practice, as you are ignoring the third dimension (z). Look which ones
>> do proper 3D. coloc2 (in Fiji), is a good one. The object-based Moore
>> plugin (
http://crg.ubc.ca/moore/) looks also good, although I haven't
>> tried it myself. If the structures you're looking at are round(ish),
>> then I must say that I have been very happy with the algorithm we have
>> produced. This will also give you a clear picture depending on the
>> z-level (e.g. if the structures colocalise more outside or inside the
>> nucleus) and the distances between the structures. I don't have the
>> paper here, but can send it to you tomorrow; the plugin requires
>> Matlab. Maybe, if it strongly depends on the z-level, it may be a
>> good-enough start to look at single planes from different z-levels
>> (using a 2D approach, e.g. Manders or Pearsons) and show that they are
>> very different.
>> Colocalisation is a tricky subject - I've ended up spending a lot more
>> time on it than I ever thought I would.
>>
>> Let me know if this helps; if not, just get back to me, I'm now back at
>> work.
>>
>> With kind regards
>>
>> Philippe
>>
>> On Thu, May 9, 2013 at 1:24 PM, Marjan Gharagozloo <
[hidden email]>
>> wrote:
>>> Dear Philippe,
>>>
>>> Thank you so much for your reply. I'm looking for RNA translocation
>>> inside nucleus. The RNA has been labeled with Cy3 and Nuclei are
>>> stained with Draq5. This is my first confocal experience and I'm
>>> really confused! My problem is that the quantity of colocalization is
>>> changing by different focal plane. So, I tried looking at Z-stacks to
>>> find something reliable. I used Zeiss software for Colocalization (LSM
>>> 710), I got some data but I'm not happy with them because I'm not sure
>>> if they are showing the correct amount of colocalization. Then, I
>>> checked imageJ, actually I found Many ImageJ software online
>>> (Imagej2X, WCIF ImageJ, and just ImageJ). Also I downloaded Huygens
>>> and Volocity 3D, but I couldn't find something related to my need.
>>>
>>> Could you kindly tell me which one is more relevant to my case? Any
>>> suggestion would be highly appreciated.
>>>
>>>
>>> Bests
>>> Marjan
>>>
>>>
>>> On 5/8/13, phil laissue <
[hidden email]> wrote:
>>>> *****
>>>> To join, leave or search the confocal microscopy listserv, go to:
>>>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>>>> *****
>>>>
>>>> Hi Marjan,
>>>>
>>>> more information would be needed to understand the problem, but I'm
>>>> pasting in a few references below. If you have a z-stack, the best
>>>> option is to do a 3D colocalisation analysis by including all focal
>>>> planes (hopefully you acquired using Nyquist-Shannon reconstruction
>>>> theorem).
>>>>
>>>> (from an earlier post):
>>>> By no means a comprehensive list, but in my humble opinion some of the
>>>> most user-friendly discussions/approaches. Really depends on the
>>>> structures in question, there's not one single approach that works
>>>> best.
>>>>
>>>> pixel-based:
>>>>
http://www.ncbi.nlm.nih.gov/pubmed/23026999>>>> diva-portal.org/smash/get/diva2:563664/FULLTEXT02
>>>> coloc2:
>>>>
http://fiji.sc/Colocalization_Analysis>>>>
http://www.ncbi.nlm.nih.gov/pubmed/21209361>>>>
http://www.ncbi.nlm.nih.gov/pubmed/15189895>>>> object-based:
>>>>
http://www.ncbi.nlm.nih.gov/pubmed/20858446>>>>
http://crg.ubc.ca/moore/>>>>
http://www.ncbi.nlm.nih.gov/pubmed/19746416>>>>
>>>>
http://www.ncbi.nlm.nih.gov/pubmed/23381680>>>> (happy to send you reprint and matlab code)
>>>>
>>>> Also worth checking out:
>>>>
http://www.ncbi.nlm.nih.gov/pubmed/17210054>>>>
http://www.ncbi.nlm.nih.gov/pubmed/22086768>>>>
>>>> Hope this helps. Kind regards
>>>>
>>>> Philippe
>>>>
>>>> ______________________________
>>>> Philippe Laissue, PhD, Bioimaging Manager
>>>> School of Biological Sciences, Room 4.17
>>>> University of Essex, Colchester CO4 3SQ, UK
>>>> (0044) 01206 872246 / (0044) 07842 676 456
>>>>
[hidden email]
>>>> privatewww.essex.ac.uk/~plaissue
>>>>
>>>> On Wed, May 8, 2013 at 5:44 PM, Marjan Gharagozloo <
[hidden email]>
>>>> wrote:
>>>>> *****
>>>>> To join, leave or search the confocal microscopy listserv, go to:
>>>>>
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy>>>>> *****
>>>>>
>>>>> Dear All,
>>>>>
>>>>> I'd like to study colocalization of Draq5 and Cy3 in Nucleus.
>>>>> Actually, this topic is totally new to me and I've read many articles
>>>>> and
>>>>> manuals to find a way for quantifying colocalization. However, I don't
>>>>> know how I can analyze colocalization in my Zstack images. I know by
>>>>> changing focal plane, I'll get different results and quantities. It's
>>>>> kind of you if send me some information to do this analysis.
>>>>>
>>>>> Many thanks and best regards:
>>>>> Marjan
>>>>>
>>>>> --
>>>>> Best regards:
>>>>>
>>>>> Marjan Gharagozloo (PhD)
>>>>> Postdoctoral Fellow
>>>>> University of Waterloo
>>>>> School of Pharmacy, PHR3002
>>>>> 10 Victoria St S, N2G 2B2
>>>>> Kitchener, ON, Canada