intensity measurements
Locked
|
hello friends,
We are interested in measuring the intensity in immnohistochemical slides (DAB stained). Could anybody suggest some software or a method to quantitate DAB intensity in tissue sections. Secondly,can these measurements be made using metomorph, ImageJ or Axiovision as these are the softwares available with us. Thanks Vaishali Kailaje Scientific Assistant, Tata Memorial Centre, Advanced Centre for Treatment Reasearch Education in Cancer Kahrghar, Navi Mumbai, India. |
|
|
Hi Vaishali,
The method I find the most reproducible and satisfactory is to separate the color channels and then perform intensity measurements in the channel that best separates your specific stain. Basically, you analyze your image as if the DAB staining were a fluorescence (intensity) channel. You can find a description and example in Pham et al., Diagnostic Pathology 2007, 2:8, Quantitative image analysis of immmunohistochemical stain using a CYMK color model. In this example, they convert their images to CYMK and analyze the yellow channel, but basically you want to pick a channel that maximizes the intensity of your strain and minimizes the intensity of the background. You could do that with RGB, or other color separation methods. This approach is easy to implement with imagej and most other image analysis software. You may have to invert your image so that your intensity measurements are positive (that is stain higher than background). The other approach is to try to separate colors using a color thresholding method, for example with the "threshold color" plugin of imageJ. I find this more difficult to work with and more unreliable, generally. good luck, Julio. -- Julio Vazquez Fred Hutchinson Cancer Research Center Seattle, WA On Jun 1, 2009, at 3:51 AM, vaishali kailaje wrote:
|
|
|
In reply to this post by vaishali kailaje
1. several of the manufacturers who sell digital pathology systems and/or image analysis software now have tools to separate and analyze color images such as those obtained from IHC chromogenic stains. Some may actually do a pretty good job at separating specific color ranges or hues. The problem here is that often the analysis has to be limited to measuring the stained area fraction, since in a color image, the "intensity" is a somewhat meaningless notion (you get the average of three values, such as red, green, and blue), and very different colors or hues may give similar "intensity" values. That's why separating color channels first and picking the one that's most representative is useful. In many cases though, measuring the area fraction that has the desired color range may be just fine. It all depends on the information that is most relevant for your experiment. 2. You also need to ask yourself whether DAB and other chromogenic stains are quantitative. You can generally distinguish pale from medium from dark staining, but the "dynamic range" is most probably more limited than, for instance, fluorescent staining. Doing some staining intensity calibrations would certainly be a good idea, to see how linear and/or quantitative this approach is. -- Julio Vazquez Fred Hutchinson Cancer Research Center Seattle, WA == On Jun 1, 2009, at 3:51 AM, vaishali kailaje wrote:
|
|
|
Check the following paper
Brey et al. Automated selection of DAB-labeled tissue for immunohistochemical quantification. J Histochem Cytochem (2003) vol. 51 (5) pp. 575-84 Best, Gil Palchik -- There are people who fight one day and are good... There are those who fight one year and are better... There are people who fight many years and are very good... But there are those who fight their entire lives: they are indispensable... Bertolt Brecht (1898-1956) How can it be that mathematics, being after all a product of human thought which is independent of experience, is so admirably appropriate to the objects of reality? Is human reason, then, without experience, merely by taking thought, able to fathom the properties of real things? Albert Einstein (1879-1955) On Jun 1, 2009, at 12:31 PM, Julio Vazquez wrote:
|
|
|
In reply to this post by Julio Vazquez
In terms of quantitation, surely the best result is to
forget RGB imaging altogether. From what I recall DAB gives a browny-black
colour, so a greyscale image should do fine. But I do echo Julio's doubts
about quantitation. If you are working in fluorescence, rather than
picking one RGB channel the best thing is to get a filter (or set a spectral
range) which matches your fluorochrome and excludes anything else, then again
record an image of just that channel.
Guy
Optical Imaging Techniques in Cell Biology From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Julio Vazquez Sent: Tuesday, 2 June 2009 2:32 AM To: [hidden email] Subject: Re: intensity measurements 1. several of the manufacturers who sell digital pathology systems and/or
image analysis software now have tools to separate and analyze color
images such as those obtained from IHC chromogenic stains. Some may
actually do a pretty good job at separating specific color ranges or hues. The
problem here is that often the analysis has to be limited to measuring the
stained area fraction, since in a color image, the "intensity" is a somewhat
meaningless notion (you get the average of three values, such as red, green, and
blue), and very different colors or hues may give similar "intensity" values.
That's why separating color channels first and picking the one that's most
representative is useful. In many cases though, measuring the area fraction that
has the desired color range may be just fine. It all depends on the information
that is most relevant for your experiment.
2. You also need to ask yourself whether DAB and other chromogenic
stains are quantitative. You can generally distinguish pale from medium from
dark staining, but the "dynamic range" is most probably more limited than, for
instance, fluorescent staining. Doing some staining intensity calibrations would
certainly be a good idea, to see how linear and/or quantitative this approach
is.
--
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA
== On Jun 1, 2009, at 3:51 AM, vaishali kailaje wrote:
Internal Virus Database is out-of-date. Internal Virus Database is out-of-date. |
|
|
Please remove me from the mailing list.
Klara Balint Postdoctoral Fellow Dr. George Coukos` laboratory University of Pennsylvania 431 Curie Blvd, BRB 2/3, Rm 1331 Philadelphia, PA 19130 Phone: 215-573-4897 Fax: 215-573-7627 Email: [hidden email] ----- "Guy Cox" <[hidden email]> wrote: > In terms of quantitation, surely the best result is to forget RGB > imaging altogether. From what I recall DAB gives a browny-black > colour, so a greyscale image should do fine. But I do echo Julio's > doubts about quantitation. If you are working in fluorescence, rather > than picking one RGB channel the best thing is to get a filter (or set > a spectral range) which matches your fluorochrome and excludes > anything else, then again record an image of just that channel. > > Guy > > > > Optical Imaging Techniques in Cell Biology > by Guy Cox CRC Press / Taylor & Francis > http://www.guycox.com/optical.htm > ______________________________________________ > Associate Professor Guy Cox, MA, DPhil(Oxon) > Electron Microscope Unit, Madsen Building F09, > University of Sydney, NSW 2006 > ______________________________________________ > Phone +61 2 9351 3176 Fax +61 2 9351 7682 > Mobile 0413 281 861 |
|
|
In reply to this post by Guy Cox
A quick Google search on "DAB staining in optical density" turned up the following link as the first hit:
http://www.jhc.org/cgi/reprint/42/8/1143.pdf A 1994 article comparing Nickle enhanced DAB staining with Flow Cytometry. The main thing to keep in mind when analyzing images of DAB, in my opinion, is that DAB (whether Nickle enhanced or not) is a light ABSORBER, and so MUST be imaged in transmission. When imaged in transmission, you should then use an Optical Density (OD) calibration for your pixel values. OD varies between 0 (white) to 3 (black). The best way to calibrate an image is to include true black (e.g. aluminum foil) and true white (no sample, cells, dust, etc.) in the sample image. When this is not possible, the next best solution is to image these conditions seperate from the sample, but during the same session on the same scope. As mentioned by Guy Cox, I would strongly recommend against using a color camera for DAB analysis. The color filters built into any color camera will muck with the OD response of the CCD! Use a monochrome camera and pick the exposure time (or integration time) to get a histogram that is not saturated at eaither end. I hope these comments are useful! Chris Tully Chris Tully Microscopy and Image Analysis Expert [hidden email] 240-888-1021 http://www.linkedin.com/in/christully 2009/6/1 Guy Cox <[hidden email]>
|
|
|
- The one issue I have with using a monochrome camera to image DAB stained slides is that one loses all color information. If the sample is counterstained with fort example a blue dye, it's impossible to tell dark brown features (DAB) from dark blue features. Color cameras are pretty good these days, and if you collect a color transmitted light image you still have the option later on to convert it to monochrome for intensity staining measurement, with the added benefit that you can try to extract different color channels and see which one best separates your DAB (or other) staining from your counterstain. Another benefit would be to be able to do positive cell counts: you can get a count for all cells (stained either blue or brown, for example), and then you can get a count for the positive cells (stained brown only). I'm not sure how one would do that with monochrome images.... a negative blue-counterstained cell might appear just as dark as a positive light-brown stained cell... Julio. On Jun 1, 2009, at 11:45 AM, Chris Tully wrote: A quick Google search on "DAB staining in optical density" turned up the following link as the first hit: |
|
|
In reply to this post by Julio Vazquez
Hi,
I'll follow Julio's comment n°2 by adding that DAB staining is a precipitate of the reaction product of peroxidase action on DAB in the presence of H2O2. SO you have to be carefull that it doesnt not reflect acuratelly the number of antibody binding sites but rather it reflects the concentration of DAB and its access to the preoxidase molecule (which can vary a lot if your antigen is not on the surface). Also the reaction is very dependant on the activity of H2O2 which in turn is affected by light, age and god knows what. So it is much more tricky to achiave a staining in a quantitatively reproducible manner than Fluorescence (which is already not that straight forward!) Good luck Eric Eric Scarfone, PhD, CNRS, Center for Hearing and communication Research Department of Clinical Neuroscience Karolinska Institutet Postal Address: CFH, M1:02 Karolinska Hospital, SE-171 76 Stockholm, Sweden Work: +46 (0)8-517 79343, Cell: +46 (0)70 888 2352 Fax: +46 (0)8-301876 email: [hidden email] http://www.ki.se/cfh/ ----- Original Message ----- From: Julio Vazquez <[hidden email]> Date: Monday, June 1, 2009 6:29 pm Subject: Re: intensity measurements To: [hidden email] > > I should add a couple of things: > > 1. several of the manufacturers who sell digital pathology systems > and/or image analysis software now have tools to separate and > analyze > color images such as those obtained from IHC chromogenic stains. > Some may actually do a pretty good job at separating specific color > > ranges or hues. The problem here is that often the analysis has to > be > limited to measuring the stained area fraction, since in a color > image, the "intensity" is a somewhat meaningless notion (you get > the > average of three values, such as red, green, and blue), and very > different colors or hues may give similar "intensity" values. > That's > why separating color channels first and picking the one that's most > > representative is useful. In many cases though, measuring the area > fraction that has the desired color range may be just fine. It all > depends on the information that is most relevant for your experiment. > > 2. You also need to ask yourself whether DAB and other chromogenic > > stains are quantitative. You can generally distinguish pale from > medium from dark staining, but the "dynamic range" is most probably > > more limited than, for instance, fluorescent staining. Doing some > staining intensity calibrations would certainly be a good idea, to > see how linear and/or quantitative this approach is. > > > > -- > Julio Vazquez > Fred Hutchinson Cancer Research Center > Seattle, WA > > http://www.fhcrc.org > > == > > > On Jun 1, 2009, at 3:51 AM, vaishali kailaje wrote: > > > hello friends, > > We are interested in measuring the intensity in > immnohistochemical > > slides (DAB stained). > > Could anybody suggest some software or a method to quantitate DAB > > > intensity in tissue sections. > > Secondly,can these measurements be made using metomorph, ImageJ > or > > Axiovision as these are the softwares available with us. > > > > Thanks > > > > Vaishali Kailaje > > Scientific Assistant, > > Tata Memorial Centre, > > Advanced Centre for Treatment Reasearch Education in Cancer > > Kahrghar, > > Navi Mumbai, > > India. > > > > > > > > > > > > |
|
|
In reply to this post by Chris Tully
A very good solution to this issue was published last year.
Journal of Histochemistry and Cytochemistry 56 (4): 313-328, 2008 Multiple Immunoenzyme Staining: Methods and Visualizations for the Observation With Spectral Imaging by Chris M. van der Loos Dick ----- Original Message ----- From: Chris Tully <[hidden email]> Date: Monday, June 1, 2009 2:46 pm Subject: Re: intensity measurements To: [hidden email] > A quick Google search on "DAB staining in optical density" turned up the following link as the first hit: > http://www.jhc.org/cgi/reprint/42/8/1143.pdf > A 1994 article comparing Nickle enhanced DAB staining with Flow Cytometry. The main thing to keep in mind when analyzing images of DAB, in my opinion, is that DAB (whether Nickle enhanced or not) is a light ABSORBER, and so MUST be imaged in transmission. When imaged in transmission, you should then use an Optical Density (OD) calibration for your pixel values. OD varies between 0 (white) to 3 (black). The best way to calibrate an image is to include true black (e.g. aluminum foil) and true white (no sample, cells, dust, etc.) in the sample image. When this is not possible, the next best solution is to image these conditions seperate from the sample, but during the same session on the same scope. > As mentioned by Guy Cox, I would strongly recommend against using a color camera for DAB analysis. The color filters built into any color camera will muck with the OD response of the CCD! Use a monochrome camera and pick the exposure time (or integration time) to get a histogram that is not saturated at eaither end. > I hope these comments are useful! > Chris Tully > Chris Tully > Microscopy and Image Analysis Expert <A href="javascript:main.compose('new','t=cptully@gmail.com')">> cptully@... > 240-888-1021 > http://www.linkedin.com/in/christully > 2009/6/1 Guy Cox <<A href="javascript:main.compose('new','t=guy@emu.usyd.edu.au')">guy@...>
|
|
|
In reply to this post by Julio Vazquez
Julio,
I would recommend that you consider taking a monochrome image for use in analyzing DAB density (Optical Density) and a seperate color image for distinguishing the blue counter stain. Although you can easily convert from color to monochrome in most (if not all ) imaging software, these conversions are weighted averages of the three color channels, which have nor relationship to the actual filters used on a given camera. They will be close, but if you are really trying to get accurate quantification is "close" really good enough? Perhaps the best option is a camera outfitted with a tunable LCD filter that can be pulled out of the light path. It will give you a color image when desired and a monochrome image when needed. Based on the experience I had while working for Media Cybernetics (www.mediacy.com), although different samples of a specific model of CCD seem to have very similar color response, different models of CCD from different manufacturers (or even from the same manufacturer) do have significant differences in color response. With that in mind I will once again say that if you are trying to quantitate DAB staining intensity, a monochrome camera is the way to go! Chris Tully Chris Tully Microscopy and Image Analysis Expert [hidden email] 240-888-1021 http://www.linkedin.com/in/christully On Mon, Jun 1, 2009 at 3:42 PM, Julio Vazquez <[hidden email]> wrote:
|
|
|
In reply to this post by Julio Vazquez
Well my take on this is that counterstaining is just crazy
if you are trying to do quantitative measurements in transmitted light
wide-field.
Guy
Optical Imaging Techniques in Cell Biology From: Confocal Microscopy List [mailto:[hidden email]] On Behalf Of Julio Vazquez Sent: Tuesday, 2 June 2009 5:43 AM To: [hidden email] Subject: Re: intensity measurements -
The one issue I have with using a monochrome
camera to image DAB stained slides is that one loses all color information. If
the sample is counterstained with fort example a blue dye, it's impossible to
tell dark brown features (DAB) from dark blue features. Color cameras are pretty
good these days, and if you collect a color transmitted light image you still
have the option later on to convert it to monochrome for intensity staining
measurement, with the added benefit that you can try to extract different color
channels and see which one best separates your DAB (or other) staining from your
counterstain. Another benefit would be to be able to do positive cell counts:
you can get a count for all cells (stained either blue or brown, for example),
and then you can get a count for the positive cells (stained brown only). I'm
not sure how one would do that with monochrome images.... a negative
blue-counterstained cell might appear just as dark as a positive light-brown
stained cell...
Julio. On Jun 1, 2009, at 11:45 AM, Chris Tully wrote: A quick Google search on "DAB staining in optical density" turned up the following link as the first hit: Internal Virus Database is out-of-date. Internal Virus Database is out-of-date. |
|
|
In reply to this post by vaishali kailaje
Hi Vaishali ,
Yes, MetaMorph has the image processing command, "Optical Density (Scaled)" [Process menu ... it is a dropin, so you may need to add it using mmadmin], that performs both shading correction and the -log10 conversion (the scaled part is you multiple by 100 or 1000 to end up with integer pixels values since MetaMorph does not have [nor especially needs] a floating point image type). specimenimage(s) - darkreference OD(scaled) = scalingfactor * -log10 ( --------------------------------------------------- ) whitereference - darkreference where whitereference = blank field of view (clean area(s) on microscope slide ... ideally the median [at each pixel] from a stack of 10 or more fields of view (the version of MetaMorph I have does not have a median operation in Stack Arithemtic, so use Average instead after deleting planes with ugly junk in them). Needs to be images with the slide in place, otherwise you mess up Koehler illumination. and darkreference = no light to the camera ... ideally the average of 10 or more exposures. The inside the (------) is the standard shading correction operation. In the ODscaled command the denominator needs to be pre-calculated ("shading reference"). Use offset=0. Assumption: that the microscope is in Koehler illumination - especially that the condenser NA is adjusted correctly, lenses clean, monochromatic light (for each chromogen or dye if using two or more). The condenser NA needs to be adjusted correctly and then left alone. Too small an NA will result in lousy resolution and all kinds of junk showing up from out of focus, too high a condenser NA will result in glare (upshot: do not just push the condenser NA to fully open; glare is an OD dynamic range killer). For more on MetaMorph's use: http://support.meta.moleculardevices.com/docs/ T10012 Using Optical Density (Scaled) for Densitometry [refers to version 2.5 - quaint]. The -log10 conversion can just as easily be done by exporting gray levels to Excel and then doing the calculation (no need for the "scaled" then). DAB is a scatterer, so strictly speaking, Beer-Lambert law does not apply to it. However, with internal standards, for example, cell lines expressing known quantities of your favorite antigen), Western blot or ELISA, etc, you could convert to "real life" units, such as: molecules/um^2 (ideally). I also strongly urge that you purchase a calibrated OD standard (Edmund Scientific's "neutral density step tablet" is fine, Reynard Corp, http://www.reynardcorp.com/filters-step-linear-neutral-density-filters-c-1_2-l-en.html but all these slides are 2 mm thick, or film based ones from Kodak or Stouffer.net). DAB IHC vs fluorescence: The dynamic range of DAB IHC is much less than fluorescence if you are using a scientific grade CCD camera for the detector (ex. Hamamatsu ORCA-ER). On a good day, your dynamic range in O.D. units is going to be about 0.05 to 1.2 (if operating in 8-bit range, which you should not do) or 0.05 to almost 2.0 (if 12-bit CCD, no staining to almost 1% transmission). Dynamic range of fluorescence can be huge, if your background (from all sources, but most obvious issue is from the tissue autofluorescence) is really low. The following paper makes a good case for fluorescence, but the author is cofounder (and patent co-inventor) of a fluorescence imaging method ("AQUA" score is basically an optical equivalent of unsharp masking followed by a scaling to useful 8-bit dynamic range ... company was too chicken to report in real number of molecules per um^ or per cell, and the unsharp masking operation could be replaced by a much more quantitative deconvolution algorithm with little extra acquisition or processing time): Rimm DL. What brown cannot do for you. Nat Biotechnol. 2006 Aug;24(8):914-6. Review. PMID: 16900128 Color vs monochrome camera: While in principle Beer-Lambert law refers to monochromatic light, a typical fluorescence bandpass filter is fine for OD imaging. Even the color filters in the color cameras are good enough (and pixel resolution is typically not an issue, or can use a 40x/0.75 NA lens instead of a 20x/0.75 NA lens if you want or need to). The BAD feature about color cameras is that they typically default to gamma NOT EQUAL to one. This is not a killer, just need to be aware of it. This also mars most papers that use a desktop scanner for "quantitative densitometry" of chemiluminescent Western blots (or autoradiograms). For example, the otherwise nice papers by Wu and Pollard doing "absolute quantitation" report absolutely incorrect values (probably off by no more than a factor of three, which is probably close enough for a biochemist) because they do not understand that their cheap desktop scanner is designed to make color photographs of people's faces show up ok on their monitor. Chapter 9: Counting proteins in living cells by quantitative fluorescence microscopy with internal standards. Wu JQ, McCormick CD, Pollard TD. Methods Cell Biol. 2008;89:253-73. PMID: 19118678 Counting cytokinesis proteins globally and locally in fission yeast. Wu JQ, Pollard TD. Science. 2005 Oct 14;310(5746):310-4. PMID: 16224022 If you want to use your scanner as a densitometer, buy VueScan Pro and use 16-bit TIFF output ( http://www.hamrick.com/vuescan/html/vuesc32.htm#outputrawfile). SilverFast Ai has a similar "raw' output option and is bundled with some scanners. If you have a color camera that is not linear (or a monochrome camera that is not linear!), and invest in an OD step tablet, then you can generate a calibration curve in Excel (or MetaMorph). Here is my KISS color imaging method for DAB&Hematoxylin: monochrome CCD (ex. ORCA-ER) DAPI bandpass filter cube -> DAB image channel Cy3 bandpass filter cube -> hematoxylin image channel Set up exposure time(s) so the bright background in each is in the 3000 to 3500 intensity range (on ORCA-ER the dark reference image will be around 200, so you will have ~3000 intensity units to play with. In theory should be ~3 OD, but reality is that no one will be able to tell by eye a cell with average OD of 2.0 vs 3.0 on the microscope, so the staining protocols will never be designed to do this). For color image: R = Cy3 filter cube image G = Cy3 filter cube image (or, to get funky, stick in the DAPI channel image instead) B = DAPI filter cube image If possible, replace hematoxylin with methyl blue (see Chantrain et al 2003 J Histochem Cytochem, note: we did NOT do OD imaging). If you are serious about OD imaging, make extra serial sections and do DAB only on half of them. If you are willing to try out OD vs fluorescence, you can use fluorescent tyramide as a direct replacement for DAB (money note: if you are not paying attention to details and figure out how to minimize the reagent volume, you might end up pipeting $100 of fluorescent tyramide on a single slide). You will probably need to re-titer your antibodies. See Brian Van Tine's methods paper for more on fluorescent tyramides: Simultaneous in situ detection of RNA, DNA, and protein using tyramide-coupled immunofluorescence. Van Tine BA, Broker TR, Chow LT. Methods Mol Biol. 2005;292:215-30. PMID: 15507711 Sincerely, George At 06:51 AM 6/1/2009, you wrote: hello friends, George McNamara, Ph.D. Image Core Manager Analytical Imaging Core Facility University of Miami, Miller School of Medicine Miami, FL 33010 [hidden email] [hidden email] 305-243-8436 office http://home.earthlink.net/~Pubspectra http://home.earthlink.net/~geomcnamara http://www.sylvester.org/AICF (Analytical Imaging Core Facility) |
|
|
In reply to this post by Klara Balint
Klara, I deleted you from the confocal list.
Jerry Sedgewick Klara Balint wrote: > Please remove me from the mailing list. > > Klara Balint > Postdoctoral Fellow > Dr. George Coukos` laboratory > University of Pennsylvania > 431 Curie Blvd, BRB 2/3, Rm 1331 > Philadelphia, PA 19130 > Phone: 215-573-4897 > Fax: 215-573-7627 > Email: [hidden email] > > ----- "Guy Cox" <[hidden email]> wrote: > > >> In terms of quantitation, surely the best result is to forget RGB >> imaging altogether. From what I recall DAB gives a browny-black >> colour, so a greyscale image should do fine. But I do echo Julio's >> doubts about quantitation. If you are working in fluorescence, rather >> than picking one RGB channel the best thing is to get a filter (or set >> a spectral range) which matches your fluorochrome and excludes >> anything else, then again record an image of just that channel. >> >> Guy >> >> >> >> Optical Imaging Techniques in Cell Biology >> by Guy Cox CRC Press / Taylor & Francis >> http://www.guycox.com/optical.htm >> ______________________________________________ >> Associate Professor Guy Cox, MA, DPhil(Oxon) >> Electron Microscope Unit, Madsen Building F09, >> University of Sydney, NSW 2006 >> ______________________________________________ >> Phone +61 2 9351 3176 Fax +61 2 9351 7682 >> Mobile 0413 281 861 >> > > -- Jerry (Gerald) Sedgewick Core Facility Director, Biomedical Image Processing Lab (BIPL) University of Minnesota, Department of Neuroscience 1-205 Hasselmo Hall 312 Church St. S.E. Minneapolis, MN 55455 612-624-6607 [hidden email] http://www.bipl.umn.edu Author: "Scientific Imaging with Photoshop: Methods, Measurement and Output." Rawlight.com (dba "Sedgewick Initiatives") 965 Cromwell Avenue Saint Paul, MN 55114 651-788-2261 [hidden email] http://www.rawlight.com http://www.jerrysedgewick.com --- Get FREE High Speed Internet from USFamily.Net! -- http://www.usfamily.net/mkt-freepromo.html --- |
«
Return to Confocal Microscopy List
|
1 view|%1 views
| Free forum by Nabble | Edit this page |