Re: Particle analysis: classifying with respect to size

Posted by Jacqueline Ross on Dec 17, 2007; 9:08pm
URL: http://imagej.273.s1.nabble.com/Particle-analysis-classifying-with-respect-to-size-tp3697736p3697738.html

Hi Dimiter,

Thanks for your reply. I agree that you have to get some initial data in
order to do this.

I intend to do the particle analysis with all particles initially and
then look at the frequency distribution and try to work out from that
which ones are likely to be outliers. I'm looking at cell/nuclei number
but some cells/nuclei are still too close together, therefore presenting
as one object although by eye, you can see from the morphology (and
size!) that there are actually 2 or more contributing to the particle. I
also have a good idea of what the size of the cells/nuclei are.

It would of course be easier if I can easily see clear peaks in the
frequency histogram corresponding to one, two, three particles but it is
a little more complicated than that. I could also choose to exclude
larger particles from the particle analysis initially since I do know
approximately how large the indivdual nuclei are (for example). Then I
could use Gabriel's classification plugin together with that data.

Anyway, I will try out this plugin since it does look like a good
option.

Kind regards,

Jacqui


Jacqueline Ross

Biomedical Imaging Microscopist
Biomedical Imaging Research Unit
School of Medical Sciences
Faculty of Medical & Health Sciences
The University of Auckland
Private Bag 92019
Auckland, NEW ZEALAND

Tel: [phone] Ext 87438
Fax: [phone]

http://www.health.auckland.ac.nz/biru/ 


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
Dimiter Prodanov
Sent: Monday, 17 December 2007 9:29 p.m.
To: [hidden email]
Subject: Re: Particle analysis: classifying with respect to size

Hi Jacquie,

I agree with Gabriel, without prior knowledge on the size distribution
it will be difficult to classify. With this statement you assume
unimodal distribution.

how many classes do you normally expect? I have an algorithm for
unsupervised classification in 2^n classes. But it is implemented so far
in Matlab.
----------------------
At the moment, my idea is to find out the mean (+ 2SD, etc.) using
Analyze Particles and then go into Excel to classify them as one or more
but if anyone has a better way of doing this, it would be great.

-----------------

Kind regards,

Dimiter