nimne
commented
3 years ago Thanks for the detailed report!
I think readlif is already reading that info?
All of the metadata is indeed read (and returnable as a big ElementTree), but not all of the metadata is used. This probably hits on two issues. Some issues with how channels are read identified in #19 (there is a warning for this), and access to additional metadata identified in #18.
Creating an explicit tree-like structure for these files may be a difficult problem, but would it be sufficient to build image 'names' based on the 'path'? For example, set the name to be something like B3/60/10x/B3/R1?
At the end is B/3/R1-R60 which ideally would be able to be imported together.
It should be possible to accomplish this because of some excellent work done by the aicsimageio project to import things in a consistent and easy to use format. A design limitation of this project is that it will only return a single 2D plane at a time - loading up batches of planes into an _n_d image happens using iterators or other wrappers. I'll take a look at the example images you've linked and try to figure out why they are not working correctly as a 'mosaic'.
Fortunately I'll have some time soon to clean up a few open issues here! However, it may be a few weeks!
psobolewskiPhD
This may be related to the existing issue https://github.com/nimne/readlif/issues/19 I'm also using LAS X Navigator and multi-well tile scans. These can be continuous—mosaic merges—or not. Originally I posted this to
aicsimageiobecause that's what I was using to get my image data into python, but now I understand it is a better fit here. Here's the original issue: https://github.com/AllenCellModeling/aicsimageio/issues/277 So using Navigator it's easy to do multi-well experiments, but then when importing names don't get read in properly, which makes everything confusing.Here's a link to a representative LIF: https://www.dropbox.com/s/jk9se8i1kpzqvsn/20210428_24w_L929_Ho_B2C3.lif?dl=0 Just as an example, we can look at the last section: B is row B, then B/n/ are wells in the B row. R1-R5 are actually mosaic merges, but that's not material to this issue. Same with C/ At the end is B/3/R1-R60 which ideally would be able to be imported together.
ReadMyLIFs by @mutterer does this perfectly:
It's easy to import all the regions together (this is like a mosaic, except scattered in a well). I know that is using
bio-formatsbut perhaps that's possible here, to pass on toaicsimageio?At the moment
aicsimageioignores the structure of the project, so I get many R1, etc. and can't call the scenes by name as a result (now can use indexes, so it works, but less than ideal, but without the well context, everything is more tricky).Here's the xml for this LIF big-lif.xml.zip
Here's the start of the last section of images:
<Attributes><Attribute>ExtendedMemoryBlock</Attribute><Attribute>_DCSERVER_PREVENT_MEMKILL</Attribute></Attributes><Memory Size="2764800" MemoryBlockID="MemBlock_1050" /><Children /></Element><Element Name="B3 60 10x" Visibility="1" CopyOption="1" UniqueID="a6f972e2-a82c-11eb-bc72-a4bb6dca3902"><Data><Collection ChildTypeTest="AcceptAll"><ChildTypeList /></Collection></Data> <Memory Size="0" MemoryBlockID="MemBlock_1319" /><Children><Element Name="B" Visibility="1" CopyOption="1" UniqueID="a6f972e3-a82c-11eb-bc72-a4bb6dca3902"><Data><Collection ChildTypeTest="AcceptAll"><ChildTypeList /></Collection></Data> <Memory Size="0" MemoryBlockID="MemBlock_1320" /><Children><Element Name="3" Visibility="1" CopyOption="1" UniqueID="a6f972e4-a82c-11eb-bc72-a4bb6dca3902"><Data><Collection ChildTypeTest="AcceptAll"><ChildTypeList /></Collection></Data> <Memory Size="0" MemoryBlockID="MemBlock_1321" /><Children><Element Name="R1" Visibility="1" CopyOption="1" UniqueID="a6f972e5-a82c-11eb-bc72-a4bb6dca3902"><Data><Image TextDescription=""><Attachment Name="TileScanInfo" Application="LAS AF" FlipX="0" FlipY="0" SwapXY="0"><Tile FieldX="0" FieldY="0" PosX="0.0534542067" PosY="0.0256040830" />You can see
<Element Name="B3 60 10x" Visibility="1" CopyOption="1" UniqueID="a6f972e2-a82c-11eb-bc72-a4bb6dca3902">And then Children:I'd love to be able to see those parent names, so something like: "B3 60 10x B3 R1" In an ideal world, because R1 has the "TileScanInfo" bit, the scene would be "B3 60 10x B3" and then all the R1-60 would be part of that as M values—akin to a mosaic, even though they are not touching.
I think
readlifis already reading that info? https://github.com/nimne/readlif/blob/581ce8be5a4717e56c44c91b2cec6ea31ee65978/readlif/reader.py#L534 But I'm 100% a python noob and can't quite figure out what's going on. I'm willing to help, hack, or test, but need a bit of guidance. Thanks for making a great package! 😍