particle counting in neurons

[Elizabeth Huang]

Hello,

I am looking at a GFP-tagged protein that we are expressing in neurons and that localizes into
fairly discrete puncta of various sizes. I am trying to count the number of particles that this
protein localizes to, but there is a little background fluorescence and sometimes it is difficult to
differentiate what is a particle and what isn't (for instance, a single very bright pixel might be a
particle, but a cluster of dim pixels might be background). Later I may also want to analyze
particle area. I need to play around more with the Analyze Particles function in IMAGEJ and see if it
will serve our purposes, but I was wondering if anyone had looked at something similar and had
any suggestions for how to approach the counting. I would like to find an automated method if
possible, to avoid both the tediousness of manual counting, and more importantly to eliminate
any biases I might introduce in deciding what to count and what not to count. Thank you--any
advice would be greatly appreciated!

EH

[Brooke Kelley]

It can be challenging.  Some people use threshold segmentation to count
the particles but if the background signal is sometimes close to the
object signal then it becomes difficult.  You can use convolution filters
and background subtraction to increase the resolution of the more bright
and defined objects but these solutions are not perfect and some images
will yield more accurate results than others.  Hand counting is the most
accurate.  I've dealt with a variety of types and quality of images and
often returned to a blinded hand counting method.  I hope someone else
will chime in with any ideas.  I'm always looking for new (faster &
easier) ways

Brooke Kelley


On Wed, 26 Jul 2006, Elizabeth Huang wrote:

> Hello,
>
> I am looking at a GFP-tagged protein that we are expressing in neurons and that localizes into
> fairly discrete puncta of various sizes. I am trying to count the number of particles that this
> protein localizes to, but there is a little background fluorescence and sometimes it is difficult to
> differentiate what is a particle and what isn't (for instance, a single very bright pixel might be a
> particle, but a cluster of dim pixels might be background). Later I may also want to analyze
> particle area. I need to play around more with the Analyze Particles function in IMAGEJ and see if it
> will serve our purposes, but I was wondering if anyone had looked at something similar and had
> any suggestions for how to approach the counting. I would like to find an automated method if
> possible, to avoid both the tediousness of manual counting, and more importantly to eliminate
> any biases I might introduce in deciding what to count and what not to count. Thank you--any
> advice would be greatly appreciated!
>
> EH
>

[Dimiter Prodanov]

Dear Elizabeth,

I would say that particle counting is quite brute force approach.
Selecting the correct threshold is sometimes difficult to justify.
Why don't you try some of the texture analysis methods.
they are fully automatic and reproducible.
For example - SurfaceJ, Granulometry, even simple
histogram statistics.
All of these support ROIs so you will be able to measure
exactly the neuron and none of the outer space.

best regards

Dimiter Prodanov

>Date:    Wed, 26 Jul 2006 18:41:11 -0400
>From:    Elizabeth Huang <[hidden email]>
>Subject: particle counting in neurons
>
>Hello,
>
>I am looking at a GFP-tagged protein that we are expressing in neurons an=
>d that localizes into=20
>fairly discrete puncta of various sizes. I am trying to count the number =
>of particles that this=20
>protein localizes to, but there is a little background fluorescence and s=
>ometimes it is difficult to=20
>differentiate what is a particle and what isn't (for instance, a single v=
>ery bright pixel might be a=20
>particle, but a cluster of dim pixels might be background). Later I may a=
>lso want to analyze=20
>particle area. I need to play around more with the Analyze Particles func=
>tion in IMAGEJ and see if it=20
>will serve our purposes, but I was wondering if anyone had looked at some=
>thing similar and had=20
>any suggestions for how to approach the counting. I would like to find an=
>  automated method if=20
>possible, to avoid both the tediousness of manual counting, and more impo=
>rtantly to eliminate=20
>any biases I might introduce in deciding what to count and what not to co=
>unt. Thank you--any=20
>advice would be greatly appreciated!=20
>
>EH
>
>------------------------------

_______________________________________________________________________
Dr Dimiter Prodanov, MD, Ph.D.

Neural Engineering Rehabilitation Laboratory
(Laboratoire de Génie de la Réhabilitation Neurale)
Département de Physiologie et Pharmacologie
Université catholique de Louvain
Avenue Hippocrate, 54
POBox UCL-5446 / B-1200 Bruxelles -Belgique-
Phone: [phone]
Fax: [phone]

http://www.md.ucl.ac.be/gren

[(Jessie) Ting Guo]

Dear ImageJ users,

I'm new to ImageJ. I'm wondering how to measure SNR using ImageJ? Any of
your answers and suggestions will be greatly appreciated.

Thank you very much!
Jessie