Deconvolution advice

[HHathrell]

Hi Stephan,

If you have a few minutes I was wondering if you could give me any advice on your deconvolution plugin. I have been able to use to MultiView Reconstruction App to register our datasets by using membrane junctions as interest points. However, I was am struggling to make any progress with the quality of the deconvolution. I have 3 angles in this particular dataset, and I am trying things out on a smaller bounding box for speed while I try the different parameters.

In particular, my results are very stripy. I have tried playing around with the blending parameters, but they seem to be producing identical results. I have tried using the option to illustrate the overlap of views, but I get files with only a single intensity value out.

I am also not sure of the best way to give the program a psf, and this may be part of the problem. We have beads in the current image, but they are of a similar brightness to the brightest fluorescence in the embryo, so I am unable to isolate them. I have cropped out a bead I have been providing, but I am not sure how cropped it should be.

I have attached uploaded the three I have tested with, and examples of the results.

I think the large crop is definitely too big, as I am getting a ringing around the edge of the embryo. However, with all the psfs I have tried I am getting strips, which become quite noticeable with more iterations, eventually obscuring the data completely. I am only using the psf of one angle so far, and transforming it. I should note I am not 100% sure the psf is maintaining the correct voxel size, but I am not sure how to check this. The image files have been read in with the correct voxel size from the metadata, and for the tif I have manually entered the units via Fiji->properties. I have also attached the Max projection of the transformed PSF for the large image, so you can see if it what you would expect.

I was wondering if you had any suggestions as to what I could try next? I have tried varying the Tikhonov parameter between 0.06 and 0.006, but the results do not seem to change.

If you have any advice, or can help me isolate the problem, please let me know.
Best wishes,
Holly Hathrell

(Srinivas Lab)