Three ways to upload z-stack to OMERO

[Kouichi_C_Nakamura]

I have tested three ways of uploading a Z stack image to OMERO.

1. Upload single slice as individual files into a dataset. This way, you'll get thumbnails of the slices at the left of iviewer, whihc is nice. The downside was that the response of clicking one of those thumbnails was quite slow and took 5 to 7 seconds to load an image.

2. Upload the stack as a stack into a dataset. You'll get only one thumbnail, which is a bit useless. However, the response of Z slider was quick and less than one second.

3. Upload slices as multiple files into a dataset but with the OME-XML header telling that they a part of a stack (I made them by ImageJ's ability to export OME.TIF). This way, you can get multiple thumbnails and Z slider. Again the response of Z slider was much faster.

Would it be possible to make the response of thumbnail clicking as fast as Z slider?

Also, I wonder if it is possible to implement the ability to show a thumbnail when you hover over the Z slider? YouTube has something similar.

[jburel]

Dear Kouichi

Thanks for the report

1. Each thumbnail will correspond to a separate image in OMERO. Each time the rendering engine will need to be iniatilized. That could explain why it was slower but 5-7 seconds is probably too much. Are you using your own server? What is the size of the plane and the type of data.
3. To have thumbnails implies that some planes will be imported as a separate images. Does it mean that you partially split planes in chunk: the ones to be used in the Z-stack and the ones to use as separate images?

From what you are describing, you are looking for a "plane" view of the data i.e. each plane has its own thumbnail without having to be a separate Image in OMERO.
Regarding the YouTube feature, we could certainly look at fetching some planes when the user hovers over some Z or T, generate a thumbnail on the fly and cache it.
This is a concept that Will Moore started to explore see https://gitlab.com/openmicroscopy/incub ... ero-player

Cheers
jmarie

[Kouichi_C_Nakamura]

It's your demo server (https://demo.openmicroscopy.org/), I guess in Dundee?
The used image was `organ-of-corti.tif` in ImageJ > File > Open Samples:

548 x 249 x 4 x 15 (X,Y,C,Z)

I just tried it again, but it took 6 to 7 seconds repeatedly.

For the option 3, I used `Write each Z section to a separate file` option of ImageJ > File > SaveAs > OME.TIFF. As a result, I had 15 OME.TIFF files to upload.

You're right. If each plane has thumbnails beforehand, then hover-preview on Z slider will be easier.

[jburel]

Dear Kouchi

Thanks for the information.
We are currently investigating why the demo server is slow, it should be much quicker to get the images.
I will compare the time with another server (different setup) but I expect it to be much faster

Cheers

Jmarie

[radhikas]

I normally take optical sectionsthrough the whole cell with a 63x or 100x objective, getting a stack that is 2-5 um thick and has 5-16 sections. I did publish a figure containing the images of the whole stack once, but the journal made me put it in a supplementary figure. Otherwise, I take a 10x image of the whole dorsal horn and sample 6-8 neurons in it at 63x, then make a composite figure using those images (with optical sections only through the middle of the cell). That way I can show cell localization and details inside the cell at the same time. To quantify receptor internalization we find it better to count by eye (i.e. without taking confocal images), because that way we can sample 50-100 neurons per experimental point. We do generate a lot of data that way, so it would be impractical to image everything. However, I show quantification and sample images together. The key is to validate the technique first and then use it.