Loading a 4D tif

[RpfR2000]

I am struggling to load a 4D grayscale tif (coordinates would be (y,x,z,t)). When I use the load function, the resulting array is three dimensional, with z and t combined in a single dimension. I can reshape but that requires knowledge of the number of t's or z's beforehand, which is generally impractical. Do I need to use a different library in this case? Thanks!

[tlnagy]

Hi 👋, TIFFs don't have the ability to store 4 dimensions natively so usually TIFF writers will store that information in some non-standard way, e.g. ImageJ writes it into the ImageDescription tag. I purposefully avoided adding this to TiffImages.jl as it is meant to load the data into Julia and then the powerful Julia array manipulation tools can take over. Regardless, it should be easy to get the z and t information out of the TIFF and then just reshape as you were doing.

However, it's hard for us to help you without more information about the TIFF. What program was used to write the TIFF? Can you provide the output of

using TiffImages
img = TiffImages.load("your img.tif")
first(ifds(img)) # prints out the tags in the first image plane

[RpfR2000]

Hi, thanks! I figured it was something like that. I generated the tiff using ImageJ, converting it from a .nd2 file (output from Nikon Elements software) -- unrelated, but if you know of a Julia package for reading in .nd2 files, I would love to know about it. I've considered writing one myself, but taking a look at similar Python libraries, it seems extremely painful. Regardless, I get the following output when I run the code you suggested:

first(ifds(timeseries)) = IFD, with tags:
Tag(SUBFILETYPE, 0)
Tag(IMAGEWIDTH, 642)
Tag(IMAGELENGTH, 528)
Tag(BITSPERSAMPLE, 16)
Tag(COMPRESSION, COMPRESSION_NONE)
Tag(PHOTOMETRIC, 1)
Tag(IMAGEDESCRIPTION, "ImageJ=1.53t
images=...")
Tag(STRIPOFFSETS, 1489112)
Tag(SAMPLESPERPIXEL, 1)
Tag(STRIPBYTECOUNTS, 677952)
Tag(XRESOLUTION, 0x000f4240//0x00eac027)
Tag(YRESOLUTION, 0x000f4240//0x00eac027)
Tag(RESOLUTIONUNIT, 1)
Tag(COLORMAP, UInt16[0, 0, 256, 512, 768, ...])
Tag(SAMPLEFORMAT, 1)
Tag(IMAGEJ_METADATA_BYTECOUNTS, UInt32[20, 882300, 110, 110, 110, ...])
Tag(IMAGEJ_METADATA, UInt8[73, 74, 73, 74, 105, ...])

[tlnagy]

Ah, ImageJ should write that info into the IMAGEDESCRIPTION tag, what's the output of

println(first(ifds(timeseries))[TiffImages.IMAGEDESCRIPTION])

unrelated, but if you know of a Julia package for reading in .nd2 files, I would love to know about it. I've considered writing one myself, but taking a look at similar Python libraries, it seems extremely painful.

I do not, sorry. I would imagine it would need to look like the wonderful NRRD package by Tim Holy https://github.com/JuliaIO/NRRD.jl

[RpfR2000]

It's very nondescript: just Tag(IMAGEDESCRIPTION, "ImageJ=1.53t
images=...")

Presumably I need to expand the output somehow?

[tlnagy]

Yes, sorry. It should be

println(first(ifds(timeseries))[TiffImages.IMAGEDESCRIPTION].data)

[RpfR2000]

Thanks, yes I was able to get this to work with the following:

    height, width, zt = size(timeseries)
    ImageJ_metadata = first(ifds(timeseries))[TiffImages.IMAGEDESCRIPTION].data
    slice_loc = findfirst("slices=", ImageJ_metadata)
    frames_loc = findfirst("frames=", ImageJ_metadata)
    hyperstack_loc = findfirst("hyperstack=", ImageJ_metadata)
    slices = tryparse(Int, ImageJ_metadata[(slice_loc[7]+1):(frames_loc[1]-1)])
    frames = tryparse(Int, ImageJ_metadata[(frames_loc[7]+1):(hyperstack_loc[1]-1)])
    if slices == nothing
        error("Could not parse the number of slices from the metadata")
    end
    if frames == nothing
        error("Could not parse the number of frames from the metadata")
    end
    timeseries = reshape(timeseries, (height, width, slices, frames))
    @show size(timeseries)