`.sdt` files created with OpenScan are unreadable in Fiji

[elevans]

I recently acquired FLIM data using OpenScan and micromanager. When I read the .sdt files in SPCImage, everything looks fine. When I try to use Fiji's .sdt reader in Bio-Formats I get a bunch of pixel noise in top set of pixels followed by empty data. Here's a gif of what I see (it's hard to explain):

It looks like the shape of the data is read correctly (the dimensions don't seem mangled) and transposing the data doesn't help. My guess is that the data blocks are not being read properly, while some other bits of the file type, like data shape, are. I'm using the latest version Fiji, 2.16.0/1.54g, but this problems occurs with older versions too (as far back as 2.10.0).

[marktsuchida]

One known limitation of OpenScan is that it doesn't write the Setup block of the SDT file (because its format is undocumented by BH, making it impractical to write reliably). BioFormats appears to use this block for some parameters (though not exclusively?). Most parameters of interest are recorded in the SDT file's other binary blocks.

But given that BioFormats is getting the size of the data correctly, that may not be related to what you are seeing.

One thing that might be worth trying is to run OpenScan with SDTCompression turned off. Does that make any difference?

[brp-optics]

Thanks, Mark! I prepared two compressed and two uncompressed OpenScan sdt files, using OWS3 on default settings for 1 frame and 3 frames, respectively. The "ground truth" is a pollen grain, and include it because it should correspond to the z-sum of the 256 tau slices in the ImageJ/FIJI stack.

Imaging conditions:

Sample uncompressed image in Micro-Manager:

Same image in Fiji:

A single-frame acquisition also works fine if uncompressed:

But the compressed data goes nowhere:

Test files here: OpenScanCompressionTest.zip

[marktsuchida]

Great!

Then I guess the question is whether OpenScan is leaving out (or writing incorrectly) something that BioFormats depends on when compressed. (I believe OpenScan-written compressed SDTs do open in BH software.)

Probably the place to start is to see where BioFormats makes the wrong decision when given the compressed files (code). I'm not set up to debug this immediately but can look at it later (or feel free to, of course).

[gselzer]

I spent a bit of time looking into this - it turns out that the images that you have been seeing aren't actually coming from Bio-Formats, but instead from https://github.com/scifio/scifio-lifesci!

When you try to load in those images, Bio-Formats does try to import them first, however it fails here, because sizeX is 0.

The reason sizeX is 0 is because SDTInfo can't find this X_IMG_STRING is missing from the SDT file's setup text.

As a result, that PositiveInteger constructor raises an exception, kicking us out of Bio-Formats, and eventually that SCIFIO loader tries. I didn't look too hard at it (yet), but I'm guessing it just can't do compression.

I wrote #23 to add X_IMG_STRING and Y_IMG_STRING to the setup text. Needs testing, there may be errors and/or other inclusions needed.

[brp-optics]

@gselzer @elevans spent a lot of time looking into this, and may have a private fork which fixes the issue on the Fiji end. There are, however, challenges in distributing readers for poorly documented formats as a part of bioformats.

[gselzer]

@brp-optics good to know. #23 is orthogonal to that though - my intent was to improve the writer so that we don't have to touch the reader!

[elevans]

As a result, that PositiveInteger constructor raises an exception, kicking us out of Bio-Formats, and eventually that SCIFIO loader tries. I didn't look too hard at it (yet), but I'm guessing it just can't do compression.

This is correct. I found the same logic path. When you get booted out of BioFormats the SCIFIO loader has no way to read the ZipStream in the file.

[gselzer]

@elevans so do you have a private fork of something, or have you gotten around this in any other way?

[elevans]

@gselzer I have a single commit that I was working on as an unpushed branch of scifio-lifesci (at the time I thought I had push writes and just started tinkering). I can share that with you if you want.