Parameter variation

In what ranges can I move when using different parameters for measuring a PSF and imaging samples?

We always recommend Recording Beads to distill a Point Spread Function (PSF) with exactly the same Microscopic Parameters that you will use later to image your specimens. But in some cases imaging in exactly the same conditions is difficult. The closer to the ideal of identical conditions, the better. This is quite vague, the problem is that it is very difficult to establish limits, because in many cases the "pathologies" of the microscopes manifest in "strange" ways.


Huygens can adapt the sampling of a measured PSF to a different sampling in an image when deconvolving it, but this adaptation should always be in the direction of undersampling the PSF and not the other way around. (Therefore you may acquire your beads in a oversampled image, but please never undersample it!!!) Still, post-acquisition resampling is not very recommendable even in this favorable case (going from a small to a larger sampling).
See alsothe FAQ Should the sampling density used in PSF measurement be equal to the sampling density of the specimen?.

In any case the beads should be imaged with a Sampling Density at least according to the Nyquist Rate, or better. A PSF derived from such beads will contain all the information about the imaging properties of the microscope, and can be adapted lateron to other imaging conditions that are slightly undersampled (read more at our FAQ What is the maximal voxel size at which Huygens can still do a good job?). An undersampled PSF does not contain the complete information about how an object will be imaged by microscope, and is therefore of not much use in the effort to recover the object from the image, i.e. deconvolution. This stresses the importance to use correctly sampled bead data.
For more details on sampling see Ideal Sampling .


A shift of 10 nm in the WaveLengths should not be a problem, but it has been observed that some confocal microscopes have Point Spread Functions (PSF's) very sensible to small changes in the wavelength, going from symmetrical to dramatically asymmetrical PSF's with small changes. That would require a new calibration, of course, and we need to go for very complex models to predict all that if the user wants to calculate it theoretically. (But we are working on that!!!)

Refractive index

Differences in the Refractive Index (r.i.) are not extremely important in absolute terms, what is more relevant is the Refractive Index Mismatch between the embedding media of the sample and the objective. This mismatch is more noticeable the deeper you explore your sample, and in regions close to the objective it is much less relevant.

The HuygensSoftware already has theoretical mechanisms to adapt the PSF to the sample depth based on the r.i. mismatch between the LensImmersionMedium and the SpecimenEmbeddingMedium, a frequent situation that produces SphericalAberration. We are working to use this depth-correction also with experimental PSF's, that is a little bit more complex. (When we have it, we would be capable to apply it also to adapt a PSF measured with some r.i. to be used in another very different r.i. situation, a case that is very usual). By now, we can only recommend to use embedding and immersion media with as much similar optical properties to each other as possible.

When you have the Huygens Professional, you can easily generate different theoretical PSF's based on microscopic parameters and varying the depth in the sample, and explore how they vary with some r.i. changes.

Let us made some tests: using a fixed Lens Immersion Medium of r.i. = 1.515, I created two theoretical two-photon PSF's for an objective with numerical aperture NA = 1.3, one using a Specimen Embedding Medium of r.i. = 1.45 and another with 1.47.

At depth = 0 (the external part of the sample) the mean-squared-error (MSE) between both PSF's is smaller that 2e-5, and the Pearson's correlation coefficient is r = 1.000 (see Colocalization Theory). They are very similar. At depth = 10 micrometers (µm) the asymmetry reflecting the spherical aberrations is noticeable in both PSF's, and the differences between them increase to make a MSE = 7e-5 and a r = 0.999. We are still in very good values, both PSF's are very similar despite the r.i. differences.

So the big thing does not come from using a r.i. of 1.45 or 1.47. The problem is the big mismatch between these values and the "lens" r.i. (actually, its oil embedding medium) which is 1.515. We can compare now the differences between two PSF's taking the same r.i. = 1.45 but calculating them at different depths inside the sample. We find that between 0 and 10 µm the MSE = 1e-4 and r = 0.987. They are noticeable different!!! (See an illustration for a similar situation in Spherical Aberration).

So if you used a medium of r.i. = 1.45 to measure your PSF and later your sample is embedded in, let's say, a r.i. = 1.47 medium, that is not really a problem. The problem begins when you want to measure your sample at deep Z positions using an oil lens of r.i. = 1.515. Theoretically calculated PSF's are adapted by Huygens to properly deal with this spherical aberrations. In the future, also experimental PSF will be corrected. In the meanwhile, if your microscope is very good and its PSF fits the ideal models quite well, you can solve it by using theoretical PSF's only. If your sample is not very thick, a depth-independent experimental PSF will do it as well.

But, as always, the better the image the more precise the restoration, so the final (obvious) suggestion is to try to go for a specimen embedding medium with r.i. matching that of the objective immersion medium. In the ideal case of perfect match, the PSF is totally depth independent.

Please read Spherical Aberration for important facts about its correction.

See the FAQ Is there a mismatch between the refractive index of the bead and the surrounding medium? for a comment about the beads r.i.