Deconvolution of Structured Illumination Microscopy (SIM) images from Apotome systems

This article discusses if for structured illumination microscopy (SIM) images taken with the Zeiss ApoTome (structured illumination) a measured or theoretically determined PSF in Huygens can be used.

Currently Huygens does not have a PSF generator for structured illumination microscopy, but we are considering its inclusion in the future, if we obtain the necessary documentation.

Images acquired in a structured illumnation system are not raw, and in the likely scenario that their processing algorithm is non-linear, the deconvolution would become questionable.

The Point Spread Function (PSF) can be attempted to be determined experimentally and theoretically. For the experimental PSF, we think that treating the bead images as recorded with a Spinning-disc microscope could be sufficient.

You can also try to deconvolve your data using a theoretical PSF. We think that ApoTome also has a strong widefield component (almost constant light flux along the volume), so you could introduce this component in Huygens in a theoretical spinning disc PSF by increasing the pinhole spacing (to 4 micron for instance) or opening up and setting the pinhole size to 500nm.

Both these values can be adapted in the parameter settings of your image (right-click on the image when it's loaded in the Huygens main window) and then saved as a microscopy template for re-use for simular datasets.

We are very interested in taking a look at both sample and bead images taken with an ApoTome. A description of the system (in terms of illumination grid frequency, processing algorithms and so on) is also very welcome!

Advantages and disadvantages of Structured Illumination

Structured Illumination Microscopy (SIM) has advantages and disadvantages, and it is recommendable to evaluate them in order to determine the optimal fit for the experimental necessities. Manufacturers claim that the quality of such a microscope is equivalent to a confocal microscope, but we are afraid that this is not exactly true. The sectioning capability in structured illluminisation (the ability to measure real 3D images as explained in Three Dee Microscope) is lost to some extent, and we certainly doubt that for thick samples there is much difference between a deconvolved widefield image and a similar structured illumination image. The algorithms on which it is based remove background blur, but we think that this is done effectively only for blur coming from planes that are away from the focal plane. At the end, to do real 3D sectioning there is nothing like a point-scanning confocal microscope.

Moreover, these processing algorithms may reduce the signal to noise ratio by amplifying the noise (as they are based on subtractions of different images) and for dim samples they may prove useless or incite bleaching. On the other hand, widefield deconvolution makes use of all the available light to restore the image, and signal-to-noise ratio should be better.

See more details in this reference: Method of obtaining optical sectioning by using structured light in a conventional microscope. M. A. A. Neil, R. Juskaitis, and T. Wilson. Optics Lett. 22, 1905 (1997).

In any case, for measuring high intensity, flat samples it may very well happen that the quality of the images obtained with structured illumination is equivalent to deconvolved widefield images: both will have a resolution in XY planes comparable to confocal microscopes, and structured illumination can provide that in a faster way than deconvolution. But it would never be able to produce the same good contrast as a real confocal, and resolution along the optical axis will be definitely worse. If bleaching is also a problem, a widefield microscope is also better, while the relative slower speed of deconvolution can be diminished significantly by using the GPU acceleration in Huygens.