ImageJ

two color live cell imaging- ratio data

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

5 messages Options

Knecht, David

two color live cell imaging- ratio data

We are trying to sort out a localization issue for one of our GFP
probes. The localization is relatively subtle relative to
cytoplasmic background so we want to be sure it is correct since
unfused probes are not uniform either. We have co-expressed our GFp-
fusion and mRFP (cytoplasmic probe) in the same cells to address this
question.
1. WHat is the best way to handle this data? It seems this should
be obvious, but I am struggling with it. FIrst choice is to simply
make a red-green or cyan-magenta merge, but that requires balancing
of the two channels and I think it is easy to get fooled trying to
look for color differences. Second, one could background subtract
(green-red) but that requires a careful balancing of the two channel
intensities plus may generate negative numbers. Alternatively, we
could do a ratio (green/red), but that seems to ask a slightly
different question. Ratio seems right to me as a way of highlighting
the relative difference in localization between the two probes.
2. Assuming one does a ratio (I am using Ratio Plus), the numbers
come out in the 0-5 range as 32 bit data. How is the result handled
in ImageJ to generate the kind of color image that I am used to
seeing from ratio data. Is it a special LUT or do you mathematically
bring the numbers to a 255 scale and then apply a normal LUT.
Thanks- Dave

Dr. David Knecht
Department of Molecular and Cell Biology
U-3125
91 N. Eagleville Rd.
University of Connecticut
Storrs, CT 06269
860-486-2200
860-486-4331 (fax)

Dimiter Prodanov

Re: two color live cell imaging- ratio data

Dear David,

I had very similar problem of the background/foreground filtering in my
fluorescence images. So what I did was to subtract the "noisy" channel
from the "signal" channel after intensity adjustment.
You can check my ChannelFilter plugin for that; If you are interested
I can send you my conference paper as well.

best regards

Dimiter

>------------------------------
>
>Date: Mon, 26 Jun 2006 10:09:29 -0400
>From: David Knecht <[hidden email]>
>Subject: two color live cell imaging- ratio data
>
>We are trying to sort out a localization issue for one of our GFP
>probes. The localization is relatively subtle relative to
>cytoplasmic background so we want to be sure it is correct since
>unfused probes are not uniform either. We have co-expressed our GFp-
>fusion and mRFP (cytoplasmic probe) in the same cells to address this
>question.
>1. WHat is the best way to handle this data? It seems this should
>be obvious, but I am struggling with it. FIrst choice is to simply
>make a red-green or cyan-magenta merge, but that requires balancing
>of the two channels and I think it is easy to get fooled trying to
>look for color differences. Second, one could background subtract
>(green-red) but that requires a careful balancing of the two channel
>intensities plus may generate negative numbers. Alternatively, we
>could do a ratio (green/red), but that seems to ask a slightly
>different question. Ratio seems right to me as a way of highlighting
>the relative difference in localization between the two probes.
>2. Assuming one does a ratio (I am using Ratio Plus), the numbers
>come out in the 0-5 range as 32 bit data. How is the result handled
>in ImageJ to generate the kind of color image that I am used to
>seeing from ratio data. Is it a special LUT or do you mathematically
>bring the numbers to a 255 scale and then apply a normal LUT.
>Thanks- Dave
>
>Dr. David Knecht
>Department of Molecular and Cell Biology
>U-3125
>91 N. Eagleville Rd.
>University of Connecticut
>Storrs, CT 06269
>860-486-2200
>860-486-4331 (fax)
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 17:46:50 +0200
>From: Johannes Schindelin <[hidden email]>
>Subject: Re: image capture
>
>Hi,
>
>On Mon, 26 Jun 2006, Gabriel Landini wrote:
>
> > I thought it would be also nice to make it able to send key presses too,
> > but I am struggling with the last bit: how to convert a string into
> > KeyCode...
>
>I have not tested it, but this snippet could work:
>
>-- snip --
>import java.awt.event.KeyEvent;
>
>[...]
>
>String toSend = "ABC-abc";
>KeyEvent e = new KeyEvent(null,
>for (int i = 0; i < toSend.length(); i++) {
> char c = toSend.charAt(i);
> e.setKeyChar(c);
> int keyCode = e.getKeyCode();
> robot.keyPress(keyCode);
> robot.keyRelease(keyCode);
>}
>-- snap --
>
>Hope this helps!
>Dscho
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 19:37:02 +0200
>From: Albert Cardona <[hidden email]>
>Subject: Re: image capture
>
>Gabriel,
>
>Yes you can have a plugin running on its own thread that never returns (never
>ends), and either use the Robot class to make clicks (your mouse pointer will
>move too, which is undesirable), or directly execute the button -just capture
>it by exploring down the Container classes. For
>example, from the WindowManager
>get the window of the video, then lists its components, then for each one that
>is a container (JPanel, etc) list its components, et cetera until you find the
>button. Then "doClick()" on the button (if it's a JButton). It all may be
>easier by creating a MouseEvent with the button
>as the object and passing it to
>whatever object is the MouseListener (this is what java does internally
>anyway).
>
>Hope it helps.
>
>Albert
>
>--------------------------------------------------------------------
>This message was sent using Webmail@INI: https://webmail.ini.ethz.ch
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 19:28:42 +0100
>From: Gabriel Landini <[hidden email]>
>Subject: Re: image capture
>
>On Monday 26 June 2006 18:37, Albert Cardona wrote:
> > Yes you can have a plugin running on its own thread that never returns
> > (never ends), and either use the Robot class to make clicks (your mouse
> > pointer will move too, which is undesirable), or directly execute the
> > button -just capture it by exploring down the Container classes.
>
>Thanks, yes I am trying via the Robot class.
>
> > For
> > example, from the WindowManager get the window of the video, then lists its
> > components, then for each one that is a container (JPanel, etc) list its
> > components, et cetera until you find the button. Then "doClick()" on the
> > button (if it's a JButton). It all may be easier by creating a MouseEvent
> > with the button as the object and passing it to whatever object is the
> > MouseListener (this is what java does internally anyway).
>
>I will have a look at this. I think that the image capture plugin is just a
>wrapper of their image capture which I guess it is written in C or C++ (it
>looks the same, although some functions are greyed out), so that may not
>work.
>Anyway, I will post the robot when it is finished because I think it could be
>quite handy.
>
>Thanks again,
>
>Gabriel
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 13:39:12 -0600
>From: Owen Lockerbie <[hidden email]>
>Subject: use of imagej to count cells?
>
>does anyone know if image j is able to count positive vs negative =
>stained cells in an immunohistochemical image? - i've attached below a =
>.jpeg image as an example of tumor cells staining positive (brown) vs =
>non-staining cells (blue). it would be great if the program can be used =
>to mark and record # of positive cells vs negative cells instead of =
>laborious and error -prone manual counting
>=20
>in the help menu of the program there doesn't seem to be a counting =
>tool, but maybe i'm just not finding it?
>=20
>any help in using the program for this purpose would be greatly =
>appreciated
>=20
>many thanks!
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>=20
> =
><https://webmail.uchsc.edu/exchange/Owen.Lockerbie/Sent%20Items/No%20Subj=
>ect-17.EML/Ki67%236-2%20061206.jpg/C58EA28C-18C0-4a97-9AF2-036E93DDAFB3/K=
>i67%236-2%20061206.jpg?attach=3D1>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>
>________________________________
>
>From: Lockerbie, Owen
>Sent: Mon 6/26/2006 1:34 PM
>To: [hidden email]
>Subject:=20
>
>
>does anyone know if image j is able to count positive vs negative =
>stained cells in an immunohistochemical image? - i've attached a .jpeg =
>image as an example of tumor cells staining positive (brown) vs =
>non-staining cells (blue). it would be great if the program can be used =
>to mark and record # of positive cells vs negative cells instead of =
>laborious and error -prone manual counting
>=20
>in the help menu of the program there doesn't seem to be a counting =
>tool, but maybe i'm just not finding it?
>=20
>any help in using the program for this purpose would be greatly =
>appreciated
>=20
>many thanks!
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 22:21:35 +0200
>From: Vincenzo Della Mea <[hidden email]>
>Subject: Re: use of imagej to count cells?
>
>If you mean automatic evaluation, ImageJ is not doing that by itself,
>but in principle can be programmed to do it. However, it is a
>difficult task, due to the many variables involved in the staining
>process (I have a student doing a master thesis on that, and I had
>other students in the past, not with ImageJ). A little bit easier
>could be if done on tissue microarrays, because of the possibility of
>including a control core with known positivity and intensity.
>By the way, your attachment cannot be reached without a password.
>Regards,
>Vincenzo Della Mea
>
>Il giorno 26/giu/06, alle ore 21:39, Owen Lockerbie ha scritto:
>
> > does anyone know if image j is able to count positive vs negative
> > stained cells in an immunohistochemical image? - i've attached
> > below a .jpeg image as an example of tumor cells staining positive
> > (brown) vs non-staining cells (blue). it would be great if the
> > program can be used to mark and record # of positive cells vs
> > negative cells instead of laborious and error -prone manual counting
> >
> > in the help menu of the program there doesn't seem to be a counting
> > tool, but maybe i'm just not finding it?
> >
> > any help in using the program for this purpose would be greatly
> > appreciated
>
>* Vincenzo Della Mea
>* Medical Informatics, Telemedicine and Ehealth Lab
>* University od Udine, Italy
>* http://mitel.dimi.uniud.it/ - http://www.eslide.net
>
>------------------------------
>
>End of IMAGEJ Digest - 25 Jun 2006 to 26 Jun 2006 (#2006-174)
>*************************************************************

_______________________________________________________________________
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: 00-322-764 5596
Fax: 00-322-764 9422

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

Knecht, David

Re: two color live cell imaging- ratio data

In reply to this post by Knecht, David

A follow up question to my question. If I do 32 bit math and divide
one channel by another or by a constant, I get pixel values ranging
from roughly 0-2 (with three decimal significance). If I then apply
the r-g ratio LUT through the Image/color/edit LUT, it applies the
LUT as if the data is 0-255. Is it autoscaling the LUT to the
dataset? Is there some way of setting the scale and/or determining
what values were assigned what colors? I would like different images
to have the same scaled LUT, but they won't if it is autoscaled.
Thanks- Dave

Dr. David Knecht
Department of Molecular and Cell Biology
U-3125
91 N. Eagleville Rd.
University of Connecticut
Storrs, CT 06269
860-486-2200
860-486-4331 (fax)

On Jun 26, 2006, at 10:09 AM, David Knecht wrote:

> We are trying to sort out a localization issue for one of our GFP
> probes. The localization is relatively subtle relative to
> cytoplasmic background so we want to be sure it is correct since
> unfused probes are not uniform either. We have co-expressed our
> GFp-fusion and mRFP (cytoplasmic probe) in the same cells to
> address this question.
> 1. WHat is the best way to handle this data? It seems this should
> be obvious, but I am struggling with it. FIrst choice is to simply
> make a red-green or cyan-magenta merge, but that requires balancing
> of the two channels and I think it is easy to get fooled trying to
> look for color differences. Second, one could background subtract
> (green-red) but that requires a careful balancing of the two
> channel intensities plus may generate negative numbers.
> Alternatively, we could do a ratio (green/red), but that seems to
> ask a slightly different question. Ratio seems right to me as a way
> of highlighting the relative difference in localization between the
> two probes.
> 2. Assuming one does a ratio (I am using Ratio Plus), the numbers
> come out in the 0-5 range as 32 bit data. How is the result handled
> in ImageJ to generate the kind of color image that I am used to
> seeing from ratio data. Is it a special LUT or do you
> mathematically bring the numbers to a 255 scale and then apply a
> normal LUT.
> Thanks- Dave
>
> Dr. David Knecht
> Department of Molecular and Cell Biology
> U-3125
> 91 N. Eagleville Rd.
> University of Connecticut
> Storrs, CT 06269
> 860-486-2200
> 860-486-4331 (fax)

Wayne Rasband

Re: two color live cell imaging- ratio data

> A follow up question to my question. If I do 32 bit math and divide
> one channel by another or by a constant, I get pixel values ranging
> from roughly 0-2 (with three decimal significance). If I then apply
> the r-g ratio LUT through the Image/color/edit LUT, it applies the LUT
> as if the data is 0-255. Is it autoscaling the LUT to the dataset?
> Is there some way of setting the scale and/or determining what values
> were assigned what colors? I would like different images to have the
> same scaled LUT, but they won't if it is autoscaled. Thanks- Dave

The 256 entry LUT is scaled to the pixel value range of the image. With
contrast enhanced images, it is scaled to the display range. The two
numbers under the plot in the Image>Adjust>Brightness>Contrast tool are
the lower and upper limits of the display range. You can also get the
display range by calling the getMinAndMax() macro function.

-wayne

Boris Joffe

Re: two color live cell imaging- ratio data

In reply to this post by Dimiter Prodanov

Dear Dimiter,

I sould also be interested in your conference paper

Thank you in advance,
Boris Joffe

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]]On Behalf Of
Dimiter Prodanov
Sent: Tuesday, June 27, 2006 10:13 AM
To: [hidden email]
Subject: Re: two color live cell imaging- ratio data

Dear David,

I had very similar problem of the background/foreground filtering in my
fluorescence images. So what I did was to subtract the "noisy" channel
from the "signal" channel after intensity adjustment.
You can check my ChannelFilter plugin for that; If you are interested
I can send you my conference paper as well.

best regards

Dimiter

(never
>ends), and either use the Robot class to make clicks (your mouse pointer
will
>move too, which is undesirable), or directly execute the button -just
capture
>it by exploring down the Container classes. For
>example, from the WindowManager
>get the window of the video, then lists its components, then for each one
that
>is a container (JPanel, etc) list its components, et cetera until you find
the

>button. Then "doClick()" on the button (if it's a JButton). It all may be
>easier by creating a MouseEvent with the button
>as the object and passing it to
>whatever object is the MouseListener (this is what java does internally
>anyway).
>
>Hope it helps.
>
>Albert
>
>--------------------------------------------------------------------
>This message was sent using Webmail@INI: https://webmail.ini.ethz.ch
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 19:28:42 +0100
>From: Gabriel Landini <[hidden email]>
>Subject: Re: image capture
>
>On Monday 26 June 2006 18:37, Albert Cardona wrote:
> > Yes you can have a plugin running on its own thread that never returns
> > (never ends), and either use the Robot class to make clicks (your mouse
> > pointer will move too, which is undesirable), or directly execute the
> > button -just capture it by exploring down the Container classes.
>
>Thanks, yes I am trying via the Robot class.
>
> > For
> > example, from the WindowManager get the window of the video, then lists

its
> > components, then for each one that is a container (JPanel, etc) list its
> > components, et cetera until you find the button. Then "doClick()" on the
> > button (if it's a JButton). It all may be easier by creating a
MouseEvent
> > with the button as the object and passing it to whatever object is the
> > MouseListener (this is what java does internally anyway).
>
>I will have a look at this. I think that the image capture plugin is just a
>wrapper of their image capture which I guess it is written in C or C++ (it
>looks the same, although some functions are greyed out), so that may not
>work.
>Anyway, I will post the robot when it is finished because I think it could
be

>quite handy.
>
>Thanks again,
>
>Gabriel
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 13:39:12 -0600
>From: Owen Lockerbie <[hidden email]>
>Subject: use of imagej to count cells?
>
>does anyone know if image j is able to count positive vs negative =
>stained cells in an immunohistochemical image? - i've attached below a =
>.jpeg image as an example of tumor cells staining positive (brown) vs =
>non-staining cells (blue). it would be great if the program can be used =
>to mark and record # of positive cells vs negative cells instead of =
>laborious and error -prone manual counting
>=20
>in the help menu of the program there doesn't seem to be a counting =
>tool, but maybe i'm just not finding it?
>=20
>any help in using the program for this purpose would be greatly =
>appreciated
>=20
>many thanks!
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>=20
> =
><https://webmail.uchsc.edu/exchange/Owen.Lockerbie/Sent%20Items/No%20Subj=
>ect-17.EML/Ki67%236-2%20061206.jpg/C58EA28C-18C0-4a97-9AF2-036E93DDAFB3/K=
>i67%236-2%20061206.jpg?attach=3D1>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>
>________________________________
>
>From: Lockerbie, Owen
>Sent: Mon 6/26/2006 1:34 PM
>To: [hidden email]
>Subject:=20
>
>
>does anyone know if image j is able to count positive vs negative =
>stained cells in an immunohistochemical image? - i've attached a .jpeg =
>image as an example of tumor cells staining positive (brown) vs =
>non-staining cells (blue). it would be great if the program can be used =
>to mark and record # of positive cells vs negative cells instead of =
>laborious and error -prone manual counting
>=20
>in the help menu of the program there doesn't seem to be a counting =
>tool, but maybe i'm just not finding it?
>=20
>any help in using the program for this purpose would be greatly =
>appreciated
>=20
>many thanks!
>=20
>=20
>Owen Lockerbie, PhD
>University of Colorado Cancer Center
>Division of Medical Oncology
>Aurora CO 80010
>Tel 303 724 3875
>email : [hidden email]
>
>------------------------------
>
>Date: Mon, 26 Jun 2006 22:21:35 +0200
>From: Vincenzo Della Mea <[hidden email]>
>Subject: Re: use of imagej to count cells?
>
>If you mean automatic evaluation, ImageJ is not doing that by itself,
>but in principle can be programmed to do it. However, it is a
>difficult task, due to the many variables involved in the staining
>process (I have a student doing a master thesis on that, and I had
>other students in the past, not with ImageJ). A little bit easier
>could be if done on tissue microarrays, because of the possibility of
>including a control core with known positivity and intensity.
>By the way, your attachment cannot be reached without a password.
>Regards,
>Vincenzo Della Mea
>
>Il giorno 26/giu/06, alle ore 21:39, Owen Lockerbie ha scritto:
>
> > does anyone know if image j is able to count positive vs negative
> > stained cells in an immunohistochemical image? - i've attached
> > below a .jpeg image as an example of tumor cells staining positive
> > (brown) vs non-staining cells (blue). it would be great if the
> > program can be used to mark and record # of positive cells vs
> > negative cells instead of laborious and error -prone manual counting
> >
> > in the help menu of the program there doesn't seem to be a counting
> > tool, but maybe i'm just not finding it?
> >
> > any help in using the program for this purpose would be greatly
> > appreciated
>
>* Vincenzo Della Mea
>* Medical Informatics, Telemedicine and Ehealth Lab
>* University od Udine, Italy
>* http://mitel.dimi.uniud.it/ - http://www.eslide.net
>
>------------------------------
>
>End of IMAGEJ Digest - 25 Jun 2006 to 26 Jun 2006 (#2006-174)
>*************************************************************