Posted by
Francisco J H Blazquez on
URL: http://confocal-microscopy-list.275.s1.nabble.com/Question-about-Optical-Density-tp7584144p7584149.html
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopyPost images on
http://www.imgur.com and include the link in your posting.
*****
Thank you all, I am finding this discussion enormously instructive.
I am new to OD measurement and it will take me some time to digest all this information, George, but I really thank you very much for the time that you spared to share your experience.
Sincerely
> Francisco Blazquez
> School of Veterinary Medicine
> University of Sao Paulo
----- Mensagem original -----
> De: "George McNamara" <
[hidden email]>
> Para: "Confocal Microscopy List" <
[hidden email]>,
>
[hidden email]
> Cc: "Sathya Srinivasan" <
[hidden email]>
> Enviadas: Sábado, 22 de Agosto de 2015 2:43:35
> Assunto: Re: Question about Optical Density
> Hi Sathya,
> Excellent links - though a lot more redundent claims than data.
> Hi Francisco,
> The online document, " Using Optical Density (Scaled) for
> Densitometry "
> ftp://ftp.meta.moleculardevices.com/pub/uic/software/MM50/MetaMorph%20Extras/Application%20Notes/ODSCALE.DOC
> .
> if the above link does not work, try using google to search for:
> metamorph scaled optical density odscale.doc
> may be of use - even though it was written in 1997 (just a youngster
> compared to the Beer–Lambert–Bouguer law
>
https://en.wikipedia.org/wiki/Beer%E2%80%93Lambert_law ).
> lot's of issues - some of which I'll even get to.DAB is not so bad -
> "talk is cheap" (as in data free electronic posts), data is better -
> more on this later (no DAB data below, but I probably have more
> experience integrating DAB and other IHC slides than all the people
> who posted on the sites Sathya mentioned)..
> Biggest point for transmitted light optical density measurements of
> small objects (small relative to the field of view): "glare this the
> killer". Good references:
> Chieco P, Jonker A, De Boer BA, Ruijter JM, Van Noorden CJ. Image
> cytometry: protocols for 2D and 3D quantification in microscopic
> images. Prog Histochem Cytochem. 2013 Jan;47(4):211-333. doi:
> 10.1016/j.proghi.2012.09.001. Epub 2012 Nov 10. Review. PubMed PMID:
> 23146330.
> Jonker A, Geerts WJ, Chieco P, Moorman AF, Lamers WH, Van Noorden CJ.
> Basic strategies for valid cytometry using image analysis. Histochem
> J. 1997 May;29(5):347-64. Review. PubMed PMID: 9184850.
> Chieco P, Jonker A, Melchiorri C, Vanni G, Van Noorden CJ. A user's
> guide for avoiding errors in absorbance image cytometry: a review
> with original
> experimental observations. Histochem J. 1994 Jan;26(1):1-19. Review.
> PubMed PMID:
> 7513318.
> the same author's RMS Handbook:
> Image Cytometry
> Author/Editor(s): P Chieco; A Jonker; Cornelis J. F. van Noorden;
> Royal Microscopical Society (Great Britain)
> ISBN: 0387916180, 9780387916187
> Publisher: Springer-Verlag
> Published Place: New York
> Published/Copyright Year: 2001
> Physical Description: Illustrations included, 24 cm
> Series: Microscopy Handbooks ( Volume 46 )
> Number of pages: 116 (x, 116 pages)
> Categories:
> Cytophotometry > Handbooks, Manuals, Etc
> Imaging Systems In Biology > Handbooks, Manuals, Etc
> Cytology > Handbooks, Manuals, Etc
> information source -
http://isbnplus.com/0387916180> see also
>
https://books.google.com/books/about/Image_Cytometry.html?id=fQRtQgAACAAJ> A major point: getting correct measurements of O.D. from a light
> microscope imaging system requires complete control of the entire
> instrument. Getting reproducible data day to day if you do not
> maintain that control - good luck.
> Another major point: under normal operating conditions in a
> biomedical research lab or light microscope core facility,
> fluorescence intensity data is not quantitative either. Getting
> within 5% is considered pretty good - see Brown et al 2015:
> Brown CM, Reilly A, Cole RW. A Quantitative Measure of Field
> Illumination. J Biomol Tech. 2015 Jul;26(2):37-44. doi:
> 10.7171/jbt.15-2602-001. PubMed PMID:
> 25802488; PubMed Central PMCID: PMC4365985.
>
http://www.ncbi.nlm.nih.gov/pubmed/25802488>
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4365985/ (open access).
> Still don't believe me about routine fluorescence microscopy
> intensity data being unquantitative? Two simple tests:
> 1. zoom down (1x, 0.7x is even better) as much as possible on your
> confocal microscope. Place a fluorescent bead in the middle of the
> field of view. Acquire an image. Move the stage so the bead is just
> barely entirely in the field of view. Acquire an image. Measure the
> intensities of the bead in the two images (a ROI is better than
> brightest pixel - the latter breaks Borisy's Law: "do not rely on a
> single pixel").
> 2. Widefield or confocal: take your favorite immunofluorescence or
> GFP specimen, optimize everything, acquire an image. Position a nice
> bright object with dark neighborhood on the stage. Now acquire a
> timelapse series of the same field of view for 10 minutes, with
> continual illumination the entire time (ex. 1 second exposure time
> => 600 image frames). Measure the ROI, export to Excel, generate
> random numbers in the next column, select both the measurement and
> random columns, sort on the random numbers (I call this
> "anti-sort"), compare the first two numbers: are they practically
> identical? Unlikely! Or, graph the measurements. Or compute the mean
> and standard deviations of the measurements. Now, in this
> experiment, you know why the data turned out this way (unless you
> cheated and used the Chroma plastic slides or ArgoLight-M slide or
> equivalent). But in real life experiments, if you are futzing with
> the focus, or repositioning the stage, or leave to take a coffee
> break, while the excitation light is on, you are unlikely to equally
> photobleach every field of view. If you use transmitted light to
> focus, good luck getting exactly the same focus point every single
> time.
> * Do not trust the NAME of measurement parameters - especially
> ImageJ's lazily named parameters. Here are the Results from Fiji
> ImageJ's Ana;yze menu -> Measure command:
> # Area Mean Min Max IntDen RawIntDen
> 1 65536 208.544 96 255 13667167 13667167
> 2 65536 0 0 0 0 0
> 3 65536 255 255 255 16711680 16711680
> where "IntDen' is short for "Integrated density" (Analyze -> Set
> Measurements). To be clear - the ImageJ does NOT claim this is
> "optical density" (though I've seen publications that use the values
> as if they were O.D.) - help page:
> "Integrated Density - Calculates and displays two values: "IntDen"
> (the product of Area and Mean Gray Value ) and "RawIntDen" (the sum
> of the values of the pixels in the image or selection). "RawIntDen"
> is only available in ImageJ 1.44c or later. "IntDen" and "RawIntDen"
> values are the same for uncalibrated image. The Dot Blot Analysis
> example demonstrates how to use this option to analyze a dot blot
> assay."
> FYI - When I learned math, I was taught to calculate the Mean FROM
> the sum of the values divided by the number of values.
> Back to the data - the values in the first row are from the Fiji logo
> file (image folder). Second row is from a black (zero values) image,
> 3rd row is from a white (255 value pixels0 image. If this
> Back to OD -
> Putting a stack of 100 OD 1 filters (with air gaps in between) would
> enable you to create a specimen of 100 OD, good luck measuring
> accurately (other than taking it apart and counting each filter ...
> yes, I realize you wrote gray levels - ImageJ doesn't take what the
> numbes a). "Garbage in, garbage out". Reasonable expectations for
> optical densities are in the range of around 0 to 2 by eye or a
> camera under typical use. Chroma, Semrock and other filter companies
> routinely show their spectra of their interference filters as %T or
> O.D. with OD down to 6.0, sometimes beyond that (7.0 on a very good
> day). They are not using a microscope-camera setup for these
> measurements! Call up tech support for each and ask (1) how do they
> do the measurements on their filters, and (2) what do they think of
> the OD spectra reported by the other company.
> "If a pixel(a) has a gray value of 100 when compared with another
> pixel(b) that has a gray value of 200 we may conclude that the first
> one (a) is darker than the second one (b) and we may conclude that
> the object region corresponding to the first pixel with the gray
> value of 100 is also darker."
> Very carefully written! Most scientists would assume that 200 is 2x
> of 100, that is, in the case of light microscopes and cameras, 200
> is 2x as many photons as 100. If the camera offset (and dark current
> for long exposures) has been subtracted out, this is correct for
> Scientific CCD and Scientific sCMOS cameras. However, for most
> cameras, flatbed scanners, etc, which apply a Gamma to the data
> after the sensor and before the application software, 100 and 200
> are not linear with respect to photons. Most flatbed scanners can do
> a pretty good job, if "raw" data is output - for example, using
> www.hamrick.com VueScan Pro, with raw data output. Think this cannot
> happen in science, or in Science? See
>
http://www.ncbi.nlm.nih.gov/pubmed/16224022 for systematically
> incorrect data from a flatbed scanner -0- which usually would not be
> a problem, except their "quantitative" fluorescence microscopy data
> is "cvalibrated" against their flatbed scanner results: "
> Fluorescence measured by microscopy or flow cytometry was directly
> proportional to protein concentration measured by quantitative
> immunoblotting. ".
> Ooops:
> "That reasoning may be extended to the Integrated Optical Density
> (IOD) that is calculated multiplying the pixel mean gray value of
> the area by the area of the region of the object where the
> measurement was taken (OD)Xarea = IOD, IOD is also the sum of the
> gray values of the area. That is, the lower the IOD number, the
> darker is the Area. "
> Incorrect! You are obtaining the "mean pixel gray value" FROM the
> integrated optical density and whatever number of pixels you have
> chosen. If you are measuring a small object, in a large field of
> view that is otherwise empty, -- and have glare controlled correctly
> (as well as practical) -- the mean will vary depending on how many
> pixels away from the object you include. The bigger the region, the
> smaller the mean value ... but this is a silly thing to do since you
> could use a small number of pixels OR the entire camera field of
> view, and calculate a mean that is for practical worth, meaningless.
> The right thing to do is to report the Integrated Optical Density of
> the object -- if the rest of the field of view is empty, all those
> pixels have individual OD. values of zero, so "adding" zeros to the
> object IOD adds nothing. More practically, the values would be so
> close to zero they should be thresholded out. A useful rule of thumb
> being that an 8-bit (linear data out!) camera has a useful pixel OD
> range of about 0.05 to 1.2 (if glare is well controlled) and a
> scientific CCD or sCMOS around 0.05 to 2.0 (maybe 3.0 if you are a
> Jonker et al caliber expert).
> DAB - not really all that bad. The stuff on an IHC slide is a POLYMER
> that scatters light. Scatters a lot at short wavelength (ex. 420
> nm), not so much as 600 nm, very little at 800 nm. Some of the
> variation between manufacturer's is that some include Nickel ions or
> Cobalt ions in their DAB formulations. Use a 420 nm narrow band
> filter is ideal for DAB on a light microscope (and scientific CCD or
> sCMOS and correct glare control). 600 nm is around the absorption
> peak of typical hematoxylin (though hematoxylins can vary too). This
> means that if you acquire the right monochromatic images, do the
> right OD image conversions and calculations, you can compute:
> OD420 - OD420(Hematoxylin) ... where this is scaled from say
> OD600(Hematoxylin) ... to get the OD420(DAB) contribution.
> I worked out most of this on a CompuCyte imaging laser scanning
> cytometer while working at City of Hope National Medical Center
> (2005-2006) - hopefully that iCyte or iCys has been retired.
> CompuCyte was bought by ThorLabs if you really want one -
>
http://www.thorlabs.com/navigation.cfm?guide_id=2333> Another instrument that might be usable (and could scan an entire
> adult human brain tissue section in one click) is Tissue Scope from
>
http://www.hurondigitalpathology.com/ (be sure to pick out the laser
> lines yourself).
> It is most prudent to image the single stained controls to become
> comfortable with this. I also note a good source of monochromatic
> light are the various laser lines on a laser confocal microscope,
> with 405 nm being close enough to my preference of 420 nm (and it is
> just a preference - a relatively narrow bandpass DAPI filter would
> work, a narrow pass BV421 emission filter should be very nice).
> By the way ... Above 700 nm, few IHC reagents absorb light (since
> pathologists and histologists cannot usually see 'up there'). Many
> flatbed scanners and 35 mm film scanners can acquire a NIR channel
> image to facilitate identifying scratches and dust -- the
> Hamrick.com VueScan Pro can let you acquire those on scanners that
> have this -- see
>
http://works.bepress.com/cgi/viewcontent.cgi?article=1027&context=gmcnamara> Even simpler ways to get quantitative histolofy and IHC are to find
> -- or buy -- either:
> 1. an ASI SKY system,
>
http://www.spectral-imaging.com/products-technologies/spectral-imaging> (the Sagnac interferometer based systems, not the color camera
> ones!).
> 2. PerkinElmer/CRi nuance (the camera) or Vectra (the digital slide
> imager).
> Both have OD calculation capabilities built in (unless ASI took
> theirs out).
> you may also find of use Macville et al 2001 ...
> downloads.hindawi.com/journals/acp/2001/740909.pdf
>
http://www.ncbi.nlm.nih.gov/pubmed/11455032> There are bunch of OD spectra of IHC dyes -- and DAB -- in the XLSX
> file inside the zip file at
>
http://works.bepress.com/gmcnamara/9/> and maybe some additional links at
>
http://home.earthlink.net/~pubspectra/ (after the two pictures that
> do not show up in web browsers - they do download).
> I don't have DAB&H posted online (other than inside the PubSpectra
> XLSX file), but some may find of interest the H&E spectra at
>
http://home.earthlink.net/~geomcnamara/skyomics.htm> and for more fun,
>
http://home.earthlink.net/~geomcnamara/CrusadeBetterMicrographs.htm> Enjoy,
> George
> > > Good morning
> >
>
> > > I have a question about ImageJ measurement of optical density in
> > > photomicrographs obtained by digital cameras. As I understand the
> > > image captured by the camera CCD register the transmitted light.
> > > That should mean that the voltage (analogical signals) produced
> > > by
> > > the CCD is proportional to the brightness of the object. After
> > > the
> > > conversion of these signals to digital information, the
> > > brightness
> > > and gray shades of the object are translated into pixels values
> > > of
> > > gray that goes from 0 (black) to 255 (white).
> >
>
> > > If a pixel(a) has a gray value of 100 when compared with another
> > > pixel(b) that has a gray value of 200 we may conclude that the
> > > first
> > > one (a) is darker than the second one (b) and we may conclude
> > > that
> > > the object region corresponding to the first pixel with the gray
> > > value of 100 is also darker.
> >
>
> > > So, when we take a measurement of the mean gray level of an area
> > > of
> > > an object (translated as Optical Density – OD), be it by using
> > > ImageJ or Axiovision (Zeiss), we should conclude that the higher
> > > values correspond to a general bright object area and the lower
> > > values correspond to a darker area of the object. In other words
> > > the
> > > value of OD has an inverse correlation with the “darkness” of the
> > > area measured.
> >
>
> > > That reasoning may be extended to the Integrated Optical Density
> > > (IOD) that is calculated multiplying the pixel mean gray value of
> > > the area by the area of the region of the object where the
> > > measurement was taken (OD)Xarea = IOD, IOD is also the sum of the
> > > gray values of the area. That is, the lower the IOD number, the
> > > darker is the Area.
> >
>
> > > If we are measuring the product of an enzyme histochemical
> > > reaction
> > > or DAB deposition in a immunohistochemical DAB reaction, we may
> > > conclude that the lower the IOD, the darker is the area and
> > > consequently, the more quantity of reaction product is present
> > > the
> > > area. Of course, this relationship must be inverted if the image
> > > was
> > > originated from immunofluorescence.
> >
>
> > > Am I right? Is there something that I am not considering?
> >
>
> > > Francisco Blazquez
> >
>
> > > School of Veterinary Medicine
> >
>
> > > University of Sao Paulo
> >
>
> --
> George McNamara, Ph.D.
> Single Cells Analyst
> L.J.N. Cooper Lab
> University of Texas M.D. Anderson Cancer Center
> Houston, TX 77054
> Tattletales
http://works.bepress.com/gmcnamara/42