Re: How to define the presence/absence of an object i n immunofluorescence

Posted by Zdenek Svindrych on
URL: http://confocal-microscopy-list.275.s1.nabble.com/Two-photon-microscope-questions-tp7583010p7583031.html

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Hi Aro,
if you want to be certain that the feature is there, you perform hypothesis
testing. Your null hypothesis is that 'there is nothing' and you want to
calculate whether you can reject this hypothesis (that is, there  'must be
something' with quite high probability, e.g. 95 percent when the 2*SD rule
is followed). However, when it comes to images, it's hard to define
'nothing' and something'...

But nobody is really doing this. The biologic (e.g. cell to cell)
variability is usually the major contribution to your statistics, so what
you normally do is: measure the 'intensity' of your feature in some
consistent way (preferably relative to other intensity-based measure of your
image) on many cells from many 'control' and 'treated' samples and perform
the null hypothesis testing on these numbers. Since you have many numbers
now, you can easily calculate their standard deviations, both 'pixel image'
contribution and biological variability is accounted for without any image-
processing theories...

I believe this is how most assays work, though I'm not a biologist...

Best, Zdenek

www.kcci.virginia.edu



---------- Původní zpráva ----------
Od: AroPro <[hidden email]>
Komu: [hidden email]
Datum: 30. 11. 2014 13:42:34
Předmět: Re: How to define the presence/absence of an object in
immunofluorescence

"Thanks a lot, Jens. You mean the intensity should be >= (mean background +
2*SD of the background)? So the difference will be statistically
significant?
Is there any reference I can refer to?
Best,
Aro

> Date: Sun, 30 Nov 2014 13:24:48 -0500
> From: [hidden email]
> Subject: Re: How to define the presence/absence of an object in
immunofluorescence

> To: [hidden email]
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Hi,
>
> 2x the SD over mean background is a good rule of thumb.
>
> Jens
>
> Jens-B. Bosse
> [phone]
>
> > On Nov 30, 2014, at 12:55, PengKe <[hidden email]> wrote:
> >
> > Dear lister:
> > I'm recently puzzled by a question that I felt might have puzzled some

of you as well and I would like to receive some suggestions and opinions
from you.
> > Question: In immunofluorescence analysis, sometimes we detected objects
that showed very weak signals, so how can we decide whether the signal is
really there or not?
> > To make the question a little more specific, I will give some artificial
numbers. Let's say the background has the signal intensity of 2000 and the
potential object showed a signal intensity of 2200. The saturating intensity
of the camera is 65535. In this case can one claim the 2200 signal represent
a real object?
> > A related question might be: is there a golden standard about how much
higher a signal intensity needs to be above the background to be defined as
an object?
> > I'm looking forward and I'll be very grateful to your opinions.
> > Best wishes,
> > Aro"