CZI: add basic support for PALM data #2897
Conversation
| @@ -174,6 +174,8 @@ | |||
| private int[] tileHeight; | |||
| private int scaleFactor; | |||
|
|
|||
| private transient boolean isPALM = false; | |||
There was a problem hiding this comment.
The transient usage here is fine as the variable is only ever used during initFile which means that it wont affect memo file usage.
Unless there are plans to make further use of this variable later I would be tempted to make it local to the initFile method would remove any confusion regarding whether or not it affects serialisation. Either approach works fine though.
There was a problem hiding this comment.
Good point - cleaned up in 8252660
|
The actual code changes here look good. From the details provided it certainly seems that the new changes are all behaving as expected. The builds and tests are also green with the associated config PR. From carrying out testing with the sample files listed I had the following results: The below files show as single channel in Zen and open as a single series with Bio-Formats, the images are empty with 0 pixel count, though this is consistent in both Bio-Formats and Zen: The below images open as described with 2 channels in Zen and as 2 series in Bio-Formats. All the metadata and pixel values look to be aligned between the 2: The following image only displayed as a single channel in Zen and as single series in Bio-Formats. The pixel values and metadata matched up correctly: As mentioned in the PR description In summary the display and metadata of the files in Bio-Formats was comparable as when viewing the files using Zen (Blue or Black). |
Fixes https://trac.openmicroscopy.org/ome/ticket/12469, see also https://trello.com/c/y6KpD3sU/18-czi-tickets
I would expect tests to remain green and memo files to be unaffected. The following are new or previously untested files:
and existing files with updated configuration:
For each, compare in ZEN vs.
showinfor ImageJ with this PR. With the exception ofPalm_Source.czi, I'd expect ZEN to show either one or two channels with the same XY size. The first channel is stored at a higher resolution, but downsampled for display; if you mouse over the image, you can see that pixel values are not displayed. TheHistotab can be used to confirm pixel min/max values, otherwise it's a matter of checking whether the image looks roughly the same inshowinf/ImageJ. The second channel, if present, is stored and displayed at the same (lower) resolution; the pixel values will be shown if you mouse over the image in ZEN.showinf/ImageJ should show the two channels as being split into separate series due to the difference in XY size. The images and pixel min/max values should look the same as in ZEN, and basic metadata (physical pixel size, channel names, etc.) should match.Palm_Source.cziis a special case in that Bio-Formats detects a single channel with 100 timepoints, where ZEN shows 2 channels as described above. All 100 timepoints should show data, and that image count is consistent with the size of the file. The thumbnail in the right-hand panel of ZEN should be consistent with the images returned by Bio-Formats. My understanding is that ZEN Lite does not have true PALM support (only ZEN black does), so in my opinion ZEN is not definitive in this case and what Bio-Formats does is reasonable. If there is a usable installation of ZEN black, that be more definitive, else happy to hear other thoughts on how this file should be handled.