I would be interested myself to find out if there is a way similar to deconvolution of spatial overlap in colour terms, but in the light of the recent query of expression of fluorescent proteins I also want to comment on the issue.
To our eye a mixture of yellow and red appears orange. Depending on the mix of photons received it will be more red or more yellow. This is why in an energy transfer set-up, where the proportion of yellow and red photons changes, the colour of the cell appears to redshift. The computer screen uses red green and blue (RGB) for generating all colours of the visible spectrum. If we know what colours have been mixed to generate a certain colour we can show that as an X/Y/Z plot or to stay with the example above an x/y plot that consists of x parts of the red dye and y parts of the yellow. However, that can be misleading if suddenly we have the yellow dye changing it's colour to orange because of pH or other reasons for the change in the molecular structure. If we than still assume the resulting orange is caused by a contribution of so x red photons and y yellow photons we are going wrong.
I use a software package called methamorph and a B&W camera. I take pictures at UV/DAPI, blue/FITC and green/PE which by design of the filters have minimal overlap, so I can later on add them together or calculate pixels of each other, as I do use the same beam splitter (and ideally a plan apochromat objective) to ensure identical pixel position for all colours. The PE fluorescence is barely visible in the FITC setting but the remaining signal can be subtracted using the green excited PE image.
Regards
Gehrard
-----Original Message-----
From: Don Schaefer [mailto:send2ds@highstream.com]
Sent: 17 August 2002 00:41
To: cyto-inbox
Subject: Compensation
Hi,
Your questions were of interest to me. Mty experience is in color imaging on film and color accuracy on CRT screens and in print. If your question hasn't been answered already, I'd be glad to contribute what I have learned if I can understand your needs better. Since I have no experience in your fields of expertise, perhaps you could answer a few questions.
What is your goal in color accuracy - to represent collected wavelengths additively, or is it to merge color data collected from several screens onto one where different values may be differentiated? Is color assigned accurately by wavelength through the software, or is it subjective to the perception of each viewer and their CRT? What color is assigned to wavelengths beyond the visible spectrum? It seems that the problem as stated needs clarification so we aren't confused by the color model of the flow process, and the color model of the final output. If that can be clarified, a correlation can be established between them and the appropriate software, probably Photoshop, can be utilized intelligently to give you a known and repeatable color value for the intended output.
Don Schaefer
Boston, MA
Responding to the following:
Date: Thu, 15 Aug 2002 09:11:17 +0100
To: cyto-inbox
Subject: Compensation ... in microscopy
Hi mr,
In my experience with confocal & CCD work you cannot
compensate data collected after acquisition, although saying that
Molecular Dynamics used to sell Unix software called 'Imagespace'
that apparently/theoretically could do compensation - never tried it
myself as it seemed impossible to tell which part of the image data
needed compensating and which did not. The new generation of CLSM's
(e.g.Lecia SP2) allow collection of data of specific wavelength's
chosen through the software and if you collect the image on
sequential - basically for example collect UV excited, then 488 and
633 nm data and then merge the data. There is obviously stretch boxes
to adjust the image colours collected for each wavelength collected
but this process is very subjective and is not mathematical as in
flow, the nearest approximation is to try and remember very roughly
what the image looked like (very difficult with Cy5 as only part of
bright epitopes can be observed by eye) and adjust the colours
accordingly.
I hope this is of some help as I am more into flow than
microscopy, there is a US company called "Universal Pixel" that sells
Mac software which seemed very good, although I'm sure there are a
lot of other alternatives available. Mac's are a better platform for
doing this as a lot of PC software pixelates the images when trying
to enlarge the image.
Best Regards
Gary Warnes,
FACSLab,
Cancer Research UK
London, UK
mr wrote:-
We are starting a multicolor project by microscopy (please, don't
hurl any epithets at me: I'm a true-blue-flow guy... just dabbling a
little in the obscure arts).
Can someone recommend decent software (preferably Mac) that can do
basic image processing that can load multiple images (taken with
different filter sets), and then compensate them to give pure
fluorochrome images?
--
Dr. Gary Warnes
FACS Laboratory
Cancer Research UK
44 Lincoln's Inn Fields
London WC2A 3PX
P: 020 7269 3394
F: 020 7269 3100
mail to:gary.warnes@cancer.org.uk
Charity No: 1089464
OCTOBER IS BREAST CANCER AWARENESS MONTH
To find out more visit www.cancerresearchuk.org
This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:26:19 EST