XCOSM Frequently Asked Questions

back to xcosm home page

Questions

  1. Can I use XCOSM to deconvolve two-photon fluorescence excitation images?

  2. If my images are saturated can I still process them with the XCOSM algorithms?

  3. What wavelength should I use for the computation of the two-photon microscope PSF?

  4. Is it possible to process a 2-D image using the LLS or EM algorithms?

  5. What does "pixel size in object space" in the "Generate PSF" menu mean?

  6. When I publish results generated with XCOSM which of your papers should I reference and how to I cite the XCOSM package?

  7. Is there some type of image conversion besides the addition of a WASHU header that must be performed before an experimentally-acquired PSF may be employed for deconvolution in XCOSM?

  8. Why is there a difference between an experimentally determined PSF and one computed with XCOSM?

  9. How can I convert a series of 2-D images into a single file of a 3-D image to be processed by XCOSM?


Answers

1. Can I use XCOSM to deconvolve two-photon fluorescence excitation images?

Yes, it is possible to deconvolve two-photon fluorescence excitation microscope (TPFEM) images. Keep in mind that TPFEM images already have a good resolution and thus there is not too much room for improving the resolution. Also, it is necessary to make sure that the pixel sizes and distance between planes in the stack are not limiting the resolution. For example, for a 1.4 NA objective and illumination wavelength of about 700 nm, the diffraction limited spot is about 600nm wide and about 1.2 um long. In this case it is necessary that the pixels are smaller than 600 nm (e.g. 100 nm pixles would be very good) and that the distance between planes is smaller than 1.2 um (e.g. 0.2 um or smaller, if possible).

To deconvolve TPFEM images, XCOSM needs to use the PSF for the two photon excitation. There are two ways to calculate the TPFEM PSF. One is directly from XCOSM: click on generate PSF and fill in the necessary information. When you come to filter click other and enter the wavelength of the excitation light, not the fluorescent wavelength. When you come to the dialog box select type of microscope PSF is for, select Confocal (circular apertures). Another dialog box will appear. For Scanning optics magnification enter 1.0 for Physical distance between apertures enter 0.0, for physical apetrure size use a small number (e.g 0.001).

The other way to generate the TPFEM PSF is to use XCOSM to compute the PSF for a wide-field fluorescence microscope (i.e. non-confocal) using the wavelength of the excitation light. When this PSF is done, square it. If you don't have a program to square the PSF, you can download one from my web site

http://3Dmicroscopy.wustl.edu/~josec/DISTRIB/FORTRAN_EM/

look for the file square.c and download it. To compile use

	cc -o square square.c

To square the PSF type

	square <input_psf_file> <output_psf_file>

The program squares the values of the PSF, but does not update the min, max, and average in the header. To update those values use the utility headit that came with the XCOSM distribution.

Go to list of questions

2. If my images are saturated can I still process them with the XCOSM algorithms?

No. The algorithms were derived under the assumption that images do not have saturated pixels (neither fluorescence saturation nor detector saturation are part of the algorithms). Therefore you should never allow saturation to occur when acquiring images for processing. To avoid saturation always set the exposure time so that the maximum intensity of your acquired image is always less than the maximum intensity that your detector is able to record. If there is a dim structure of interest then AFTER deconvolution you can adjust the color map in order to visuliaze dim structures at the expense of saturating other parts of the image.

Go to list of questions

3. The PSF of a 2-photon fluorescence excitation microscope is the square of the wide-field PSF at the EXCITATION wavelength, not at the fluorescence wavelength. However, XCOSM asks for the FLUORESCENCE wavelength only. Should I just enter the excitation wavelength for that?

XCOSM asks for the excitation wavelength because the user interface was written before two-photon microscopes were commercially available. For wide field microscopes, the fluorescent wavelength defines the PSF. The confocal PSF depends on both, the excitation and fluorescence wavelengths. However, if the Stokes shift (the difference between the excitation and fluorescence wavelengths) is small, as usually is, then using only the fluorescent wavelength has a negligible effect on the PSF. For multiphoton fluorescence, the PSF is defined by the excitation light because the excitation beam is focused to a point and scanned throughout the specimen. Thus, for the computation of a multiphoton fluorescence microscope PSF you should enter the excitation wavelength.

Go to list of questions

4.  Is it possible to process a 2-D image using the LLS or EM algorithms? If so, how shall I set the parameters in the "Generate PSF" menu?

Yes, you can do that. All you have to do is set the number of planes in Z equal to 1. You still have to give the rest of the parameters as you would for a 3-D image.

Go to list of questions

5. What does "pixel size in object space" in the "Generate PSF" menu mean? Is it the size of a single element of the CCD camera or the size of a single pixel of the image acquired by the CCD camera?

Pixel size in object space is equal to the physical CCD pixel size divided by the magnification of all the lenses between the specimen and the camera. For most microscoopes, the only lens between the specimen and the camera is the objective lens. For other microscopes, there may be a additional relay lens that further magnifies the image to the camera. In confocal scanning microscopes, the scanning mechanism usually includes additional lenses. For example: if your CCD's specifications say that the well size of the CCD array is 13 micrometers and you are using a 100x  objective lens then the pixel size in object space is 0.13 micrometers. We consider this to be the pixel size in each 2-D image acquired by the CCD.

Go to list of questions

6. When I publish results generated with XCOSM which of your papers should I reference and how do I cite the XCOSM package?

It depends on which algorithm you used to process your images. Please refer to the XCOSM web page (http://www.essrl.wustl.edu/~preza/xcosm) where a list of all the algorithms and references for each algorithm are provided. XCOSM citations can be made by referring to the XCOSM web page given up.

Go to list of questions

7.  It appears that the PSFs that are generated by XCOSM do not resemble the hourglass-shape of our acquired experimental PSF which did not deconvolve the images correctly. Is there some type of image conversion besides the addition of a WASHU header that must be performed before an experimentally acquired PSF may be employed for deconvolution in XCOSM?

Our PSFs are stored 'un-centered' so that the hourglass shape is split across the 8 corners of the PSF cube. This gives the 'center point' of maximum intensity at voxel coordinate (0,0,0). Program 'shift3d' provided in the XCOSM utility programs can be used to 'un-center' or 'center' a PSF with a valid WASHU header.

Go to list of questions

8. I've been experiencing problems when using an experimental PSF for deconvolution. My experimental PSF measured from a 0.1 micrometer fluorescent bead seems to be much larger in extend than the one generated by XCOSM. Why is there a difference and why does the experimental PSF not work well?

Sometimes differences in the size of the PSF suggest that the image of the expiremental PSF may be "saturated", this means that, at the exposure time you used to collect the PSF, many of the pixels were brighter than the maximum intensity that your detector is able to record. All those pixels are then recorded at the largest unsigned number that the detector is able to record resulting in an artifactually larger in extent PSF. You should never use a saturated PSF image for deconvolution. To avoid saturation always set the exposure time so that the maximum intensity of your acquired image is always less than the maximum intensity that your detector is able to record.

Go to list of questions

9. How can I convert a series of 2-D images into a single file of a 3-D image to be processed by XCOSM?

There are different ways to generate a file with a 3-D image from a set of files with 2-D images. For example,  make sure suppose you have three files with no headers or footers just the bytes that represent the data and they are called file1, file2, file3. Then one simple way is to use

cat file1 file2 file3 > image

Or an equivalent approach is to use:

cat file1 >> image

cat file2 >> image

cat file3 >> image

Now if your 2-D image files have a header then you can use a program to remove the header first. For example, program rmfirst512.c removes a header of size 512  bytes (you can change that to whatever your header size is). Then you run the program as follows:

rmfirst512 < file1 >> image

rmfirst512 < file2 >> image

rmfirst512 < file3 >> image

Once you have the file image you need to use the utility program 'addh' that comes with the XCOSM distribution to add a header to the 3-D image. For example if the 2-D images were 128x128 then you would use

addh image 128 128 3 4

where the last number means ushort data type) if you type addh it will give you the usage and the different data types.

Go to list of questions

back to xcosm home page


Prepared by: José-Angel Conchello and Chrysanthe Preza
Comments to: preza@essrl.wustl.edu
Last modification: Mar/21//2000
©2000 José-Angel Conchello