strange behaviour of tomo3d denoiser in aretomo reconstruction

[plosana]

When trying to apply tomo3d denoise tomogram over an aretomo tomogram, it seems that some tomogram values are not being detected well by tomo3d

Still, the protocol finishes and the output tomogram is full black.
However, if we take the aretomo tomogram and run an imod-tomo preprocess to set mean to 0 and STD to 1, the tomo3d denoiser is able to denoise


Perhaps the aretomo tomogram could be normalized when stored in scipion so that the imod preprocess is not needed?

[azazellochg]

@plosana I suggest to contact AreTomo developer for this issue. I suppose the WBP scaling depends on your volume size and raw data avg/stdev.

[pconesa]

Is it a aretomo issue or tomoedd?

[azazellochg]

@plosana has to provide more information about input TS vs tomogram statistics

[pconesa]

I've run the aretomo test and then tomoeed denoising. It all worked. I also used Backprojection in aretomo just in case and all worked.

With aretomo SIRT:

00084: Statistics (avg, std, min, max): 35.843 5.654 0.035 111.735
00085: iteration number: 1 / 5 ; K (lambda): 3.39
00086: iteration number: 2 / 5 ; K (lambda): 2.43
00087: iteration number: 3 / 5 ; K (lambda): 2.22
00088: iteration number: 4 / 5 ; K (lambda): 2.08
00089: iteration number: 5 / 5 ; K (lambda): 1.97

With aretomo backprojected:
00083: Initial
00084: Statistics (avg, std, min, max): 0.001 0.068 -1.859 0.734
00085: iteration number: 1 / 5 ; K (lambda): 0.03
00086: iteration number: 2 / 5 ; K (lambda): 0.03
00087: iteration number: 3 / 5 ; K (lambda): 0.03
00088: iteration number: 4 / 5 ; K (lambda): 0.03
00089: iteration number: 5 / 5 ; K (lambda): 0.02

And they look gorgeous!!

[azazellochg]

Then maybe it's those specific input TS that need normalization