Add Zernike interpolation to feed calibration to CompressedSpotHologram

[sumiya-kuroda]

Hello,

I’m using wavefront_calibrate_zernike() with multiple calibration points (~100 grid points), as in the jupyter notebook example. However, unlike wavefront_calibrate_superpixel(), there is currently no built-in method to generate and store a global phase correction into slm.source["phase"] (which is used by set_phase() later).

What is the intended workflow for applying the wavefront correction from the Zernike calibration? My understanding is that the best approach is to bake the Zernike correction into the hologram itself. If so, should we add something like hologram.optimize() (or similar) to incorporate this correction during phase mask optimization?

[sumiya-kuroda]

After thinking about it overnight, I believe I can implement a function that takes a coordinate in ij, interpolates the coefficients from the Zernike calibration results, and compute the wavefront correction phase for that particular point, which can be added to source['phase']. However, I’m not sure how to handle more complex patterns, such as multiple target points or when FeedbackHologram was called.

[cpanuski]

Hi @sumiya-kuroda,

Thanks for your note and apologies for the delay in getting back to you. You are correct -- there is currently no way to extrapolate multidimensional calibration (a slm.source["phase"] mask for each calibration point) to other points beyond those in the original calibration pattern.

Our approach to this has been to interpolate Zernike coefficients and then input the interpolated set of Zernike vectors into a CompressedSpotHologram to produce a single correction mask for the target pattern. This is essentially a combination of the multipoint and Zernike tutorials, but we don't have a clean user API for it. I've tagged for addition but let us know if you need this functionality earlier!

[sumiya-kuroda]

I see! Thank you very much for the response! Yes, we are certainly interested in implementing/using it! I will also explore if I could manage to make it work by myself. I will keep you updated!

[sumiya-kuroda]

Hello! I had a meeting with my PhD supervisor yesterday, and we are hoping to get this calibration pipeline running sooner for our holographic optogenetics setup (which we use MBF Bioscience’s ScanImage). If you don’t mind, I may reach out with a few questions regarding the implementation and would greatly appreciate any advice or guidance.

[ichristen]

Hi @sumiya-kuroda ,

We can add 2D interpolation (where all points are in a plane) fairly easily as the code is already written.

For optogenetics, you'll probably(?) need 3D interpolation and your calibration probably(?) will be sparse. Rather than doing interpolation by triangulation (the 2D algorithm, which relies on sufficiently high calibration point density), it might be advantageous to instead fit a 3D analytic approximation (linear or quadratic) for each Zernike term of the anisoplatanism. A sparse set of points (e.g. 10-20) can perhaps give a good fit that would remove the most significant aberration terms over the plane (spherical aberration, etc). Note that this will not compensate for very local aberration, though to be honest interpolation probably won't help you for very local aberration either. The other way to do this is to do no interpolation and instead just calibrate directly at each site of interest, but that's super sample/application dependent. Would this be of interest? Thoughts?

[sumiya-kuroda]

Hi @ichristen,

Thank you very much! I see — the 3D interpolation sounds interesting and also feasible for us. In a typical experiment, the number of holograms we need per animal can range from ~10 to 1000, distributed randomly across a FOV, so calibrating spots or holograms one by one can take a very long time.

My naïve thought was that a 2D wavefront correction (i.e., done by 2D interpolation) might already be sufficient. From a biologist’s perspective, what ultimately matters is avoiding off-target excitation and achieving uniform optogenetic excitation across all stimulated cells. But of course, this probably also depends on how well (badly) aligned my microscope is in the first place.

If I want to start with 2D interpolation, where should I take a look at it first?