Huygens2 - Basics of Deconvolution

Huygens2 - Basics of Deconvolution

Due to the laws of optics, a point-like light source is imaged by the optical components of a light microscope not as a point, but rather as a more or less diffuse "cloud"-like structure, characterized by the so-called "point spread function" (PSF). The PSF describes the intensity distribution of a point light source as recorded by the optical system. If this "real" image of a theoretical point-like light source is known, this 3D intensity distribution can be used to calculate the estimated (!) intensity distribution of the light not influenced by the optical components of the microscope (i.e.the estimate of the real object).

The point spread function of the actual microscope/lens combination can be calculated by imaging the fluorescence of small beads of known properties (diameter, fluorescence distribution etc.) The point spread function depends on several microscope and sample parameters like microscope type and alignment, lens type, numerical aperture, excitation and emission wavelength, refractive indices of the lens immersion fluid and of the mounting medium etc.).

Acquisition of image stacks for deconvolution

There are many parameters to be set correctly for a successful and reliable deconvolution. Use adequate settings to record image stacks for deconvolution!

It is a good idea to use standard xyz pixelsize settings for each objective (depending on magnification and zoom), since many parameters need to be adjusted only once for deconvolution.

Please ask Günter Giese (Light Microscopy Facility) for measured and calculated PSFs.

Deconvolution is claimed to run faster when stacks have even x, y and z dimensions (e.g. 512 x 512 x 120 voxels). Dimensions can be adjusted afer recording by loading the stack into IMARIS and defining a subregion with Visualisation.Subregion. Notice that due to a bug in the subregion tool the lower limit of each of the three dimensions is set to 0 instead of 1 (as in the depthview menu). Therefore, add 1 to the lower limit of the subregion tool window! Then save this subregion by saving as an imaris image and load into HUYGENS.

Rapid deconvolution with calculated PSF

  • open a unix shell (Desktop.open unix shell)
  • login with your network-wide account
  • login to lego (term lego)
  • for rapid multiprocessing option, limit the maximum number of processors for huygens to 1,2,3 or 4
  • e.g., type setenv OMP_NUM_THREADS 2
  • start huygens2 in the background (huygens2 &)
  • in the huygens Tcl Output window: set global undo off (huOpt gundo off), or by menue
  • in the huygens Tcl Output window: set verbosity off (huOpt verb -mode noQs)
  • in the Huygens menu: load image (file open)
  • activate the desired image
  • correct parameters (Edit Parameters: x, y and z sample size, refractive index, pinhole size, channel-specific wavelengths etc.). Important: The pinhole size used by HUYGENS represents the "backprojected pinhole size", but, by IMARIS 2.7, the "digital pinhole size" is read. If you have to convert a digital pinhole size value into the backprojected pinhole size, use the translation table, but be aware of the difference between digital, physical and backprojected pinhole size (ask Günter Giese).
  • split images containing multiple channels into single channel images, delete original
  • estimate the signal to noise ratio:
    • go to a region remote from any object signal (usually the first or last image of a stack)
    • find brightest pixel in a structure (not just a bright noise pixel): this gives the "signal"
    • in a region obviously devoid of signal originationg from structures: find lowest pixel values
    • in the same region: find value of noise pixel
    • subtract lowest pixel value from noise pixl value (result: "noise")
    • divide "signal" by "noise". Take the sqare root of this result to get the S/N ratio
  • estimate background (lowest). Determination of the background by the Huygens2 program:
    • estbg-mode lowest)
  • load the psf
  • it may necessary to mirror in Z the PSF or the images stack (it depends on the scan direction)
  • scale the psf, if necessary (script scaleIm.tcl). Scaling factors of less than 1 enlarge the pixelsize (??)
  • perform mle deconvolution

Deconvolution with measured PSF

Generate your own PSF

Some initial work has to be done to generate a measured point spread function (PSF), and several issues must be taken into account not to run in trouble with wrong data. But a measured PSF represents and takes into account the individual characteristics of the lenses, of the scanner (and its alignment by service personnel), of the mounting and the immersion medium etc. You may get an already measured point spread function.

Here is a short overview; you may ask for a detailled protocol:

  • prepare bead sample: prepare a sample of small beads (e.g. TetraSpeck 200 nm from Molecular Probes) in the medium used to mount your final sample (or in a medium of the refractive index of your choice)
  • record bead images: follow the Nyquist Criterion: choose a voxel size (in x,y and z direction) not smaller than the respective resolution divided by 2.5. To calculate this voxel size, you can use the Web-Calculator from BITPLANE . Further settings: see the original Huygens2 manual
  • calculate average sphere (with Huygens2): load the beads' scans and correct the parameters (!) You may use an Excel Macro (ask G. Giese) to convert a LEICA *.exp-File into a readable format.
  • generate a (theoretical, band-limited) sphere
  • generate a PSF

Obtain a measured PSF (staff only)

Ask

man_silhouette.png
Dr. Günter Giese
Light Microscopy Facility
Phone:+49 6221 486-360Fax:+49 6221 486-325

Email: guenter.giese@mpimf-heidelberg.mpg.de

Dept. of Biomedical Optics, MPI fuer Medizinische Forschung, Jahnstr. 29, D-69120 Heidelberg, Germany

http://lightmicro.mpimf-heidelberg.mpg.de/index.html

for help. Thereupon, follow the Huygens manual for further details.

 
http://www.mpimf-heidelberg.mpg.de/facilities/light_microscopy/tutorials/deconvolution/huygens/manual
loading content